LL1-09 hindsight check: Antimicrobials as growth promoters: resistance to common sense#

Late lessons from early warnings: the precautionary principle 1896–2000 (EEA Environmental Issue Report No 22, 2001), Chapter 9, pp. 93–100. Authors: Lars-Erik Edqvist and Knud Børge Pedersen.

Checked September 2026. The general web-search budget for this session ran out before this check began. So every source below was retrieved by fetching it directly from a primary repository:

• EU law and case law: EUR-Lex and the EU Publications Office (CELLAR) API;
• US regulation and data: the US Federal Register and FDA’s own sales data;
• international bodies: WHO IRIS, WOAH and UN documents;
• national and EU surveillance: DANMAP (Denmark), Swedres-Svarm (Sweden) and ECDC reports;
• peer-reviewed papers: via Europe PMC.

Secondary sources are flagged where they are used. Where a claim could not be verified from a primary source, the text says so.

When the check resumed later on 25 September 2026, web search was available again. Several of the most recent items were then re-checked against their sources: the ECDC 2024 VRE figures, the Frederiksen et al. (2026) narasin paper, and Implementing Regulation (EU) 2026/1189. News coverage was also checked to confirm that the EU import restriction on Brazil took effect on 3 September 2026.

Overview#

What holds or has been strengthened. The chapter’s direction of travel was right, and much of its evidence has aged well.

• Removing growth promoters cut resistance in animal reservoirs. Danish surveillance recorded large, fast falls after each withdrawal:
• glycopeptide-resistant E. faecium in broilers fell from 72.7% (1995) to 5.8% (2000);
• erythromycin resistance in pig enterococci fell from about 90% to 47% and 28%;
• erythromycin resistance in Campylobacter coli from pigs fell from 68% (1998) to 24% (2003).

Sources: Aarestrup et al., 2001; DANMAP 2008. - An independent review endorsed the Danish programme. The WHO international review of Denmark (2003) found “no serious negative effects” and called the programme “very beneficial in reducing antimicrobial resistance in important food animal reservoirs”. - The virginiamycin challenge failed. The manufacturer lost its case. On 11 September 2002 the Court of First Instance dismissed Pfizer’s action (T-13/99) and a parallel action over bacitracin zinc (T-70/99). The Pfizer judgment became a leading EU authority on the precautionary principle. - Low-dose selection is now established. The view that sub-inhibitory doses cannot select for resistance (p. 94) has been decisively refuted by later laboratory work showing selection far below inhibitory concentrations. - The recommendations were adopted, slowly. The phase-out the chapter’s sources recommended was eventually adopted: - EU-wide from 1 January 2006; - in the US for medically important drugs from 1 January 2017; - endorsed in principle by the UN General Assembly in October 2024.

Since 3 September 2026 the EU has applied its ban to imported animal products as well. Its first effect was to suspend several categories of Brazilian animal products, including beef and poultry. - Colistin strengthened the general claim. The strongest later evidence that feed use of an antimicrobial affects human resistance comes from colistin in China. The plasmid gene mcr-1 was discovered in 2015, and after colistin was banned as a growth promoter in 2017, mcr-1 carriage fell in both animals and people (Wang et al., 2020).

Where the record is more complicated.

• VRE is a weaker flagship example than the chapter claimed. The chapter’s “most convincing” example was vancomycin-resistant enterococci (p. 97). It turns out to be the weakest link between animal use and clinical harm, even though the resistance mechanism and the community reservoir are confirmed.
• Genomic studies show that hospital VRE belongs mostly to hospital-adapted lineages that are distinct from livestock strains (Willems et al., 2005; Lebreton et al., 2013; Gouliouris et al., 2018).
• The US had a large hospital VRE epidemic without ever using avoparcin (Bonten et al., 2001).
• The hospital VRE burden in Europe has grown two decades after the avoparcin ban. ECDC estimates that VRE infections rose from 47,124 in 2016 to 117,866 in 2020. Vancomycin resistance stood at 16.5% of invasive E. faecium across the EU/EEA in 2024 (level with 2020), and at 13.9% in Denmark, where it has been rising for a decade.
• Swann’s “vindication” is strong on mechanism and policy, weak on measured human outcomes (p. 98).
• WHO’s 2017 guideline recommended complete restriction of growth-promoter use as a “strong recommendation” based on “low quality evidence”.
• Systematic reviews find clear reductions in resistance in animals, and only probable, unquantified effects on resistance in people.
• Sweden’s fall in use was neither simple nor all due to the ban (p. 95).
• After the 1986 ban, therapeutic group medication rose for several years.
• Up to 75% of pigs received medicated feed before better management and zinc oxide brought use down (Wierup, 2001).
• Ionophore coccidiostats stayed permitted. Research published in 2012–2026 shows one of them, narasin, can co-select for VRE in broilers.
• Denmark’s transition had real costs (pp. 95–98). The chapter says nothing about them. Post-weaning diarrhoea and therapeutic use in pigs rose after the 1998–2000 withdrawals, and antimicrobial use per kg of pig rose from 31 mg (1999) to 49 mg (2008) before later measures cut it. The critics’ predictions of lasting damage (Casewell et al., 2003; Phillips et al., 2004) were not borne out in the long-run productivity data. But the transitional problems they described were real, and WHO’s panel confirmed them.
• The drug Denmark acted to protect proved clinically marginal (p. 96). The Synercid VRE indication was later removed for failure to show clinical benefit, and newer drugs displaced it.

Details to correct.

• The court. The virginiamycin action was brought before the Court of First Instance, not the European Court of Justice (p. 96; Table 9.2).
• The Swedish figures. The tonnages on p. 95 cannot be checked directly against Sweden’s official series, which is population-corrected and omits 1985. That series supports a roughly halving from 1984 to 1996, but it also shows a post-ban rise that the chapter’s “as a result” hides.
• Evidence for the Walton rebuttal (pp. 94–95). What refuted Walton before the bans were field studies at growth-promoting doses. Laboratory proof of selection below inhibitory concentrations came only in 2011.

Annex 3 of the 2013 report (Late lessons from early warnings: science, precaution, innovation, EEA Report No 1/2013, pp. 731–734) updates this case in a short piece by Ellen Silbergeld.

• Its argument. It argues that little had changed in policy by 2013. Early adopters, it says, showed that bans did not harm productivity or animal welfare, and the case had become “a case study of intransigence in the face of overwhelming evidence”. It rests the economic argument on:
• the Perdue broiler trial (Engster et al., 2002), as reanalysed by Graham et al. (2007), which found a net loss of $0.0093 per bird from using growth promoters;
• a USDA analysis of swine farms (McBride et al., 2008).
• Where later evidence supports it. Its broad direction is consistent with later evidence (WHO 2003; Aarestrup et al., 2010).
• Where it overstates. Several statements go further than the evidence supports:
• it describes the US figure of “nearly 80 %” as antimicrobial production used “as feed additives”. FDA sales data cover all food-animal uses, and in 2015 about 38% of them were drugs not used in human medicine (mainly ionophores);
• its statement that the US had restricted “only one drug” overlooks FDA’s January 2012 prohibition of certain extralabel cephalosporin uses;
• its claim that MRSA was the single most important cause of infectious illness and death in the US in 2012 is uncited;
• “overwhelming evidence” of human harm sits uneasily with WHO’s own 2017 rating of that evidence as low quality.
• Annex 2 of the same report repeats the 2001 timeline unchanged (Table A2.9), including the garbled 1975 entry (“tolysin”).

Overall weight. Most of the evidence for “vindication” comes from Danish and Swedish surveillance institutions: DANMAP, the Danish Veterinary Laboratory and SVA. These are the authors’ own institutions or their close partners. The chapter does not disclose this.

The WHO 2003 panel provides an independent international check, and it largely confirmed the Danish account. The lessons about regulatory process carry heavy weight:

• the retreating reassurances;
• the unmeetable proof standard;
• the value of monitoring;
• the court’s endorsement of acting under uncertainty.

The lessons about quantified human harm carry less weight than the chapter implies.

Claim 1: “Substantial scientific evidence” that growth promoters contribute to resistance in humans, “most convincingly” for VRE#

Original claim (p. 97). “In the last few years substantial scientific evidence has shown that the use of antimicrobial growth promoters in food animals contributes to the problems of antimicrobial resistance in humans. This has most convincingly been shown for vancomycin-resistant enterococci.” No citations are attached. The support in the chapter is the 1994–95 VRE findings and the Danish Veterinary Laboratory’s 1995 conclusion that VRE “can be transferred to humans via the food chain” (p. 96).

Subsequent developments.

Evidence that the avoparcin–VRE link was real at the farm and community level.

• Farm-level association. A Danish retrospective cohort study found avoparcin use associated with VRE on farms. The adjusted relative risk was 2.9 (95% CI 1.4–5.9) in poultry, and 3.3 (0.9–12.3) in pigs (Bager et al., Prev Vet Med, July 1997). This was one of the studies behind the bans.
• Decline in animals after the ban. After the Danish ban, glycopeptide resistance in broiler E. faecium fell from 72.7% (1995) to 5.8% (2000). In pigs it stayed at about 20% until tylosin use fell in 1998–99, because the macrolide and glycopeptide resistance genes were “genetically linked”. It then fell to 6.0% (Aarestrup et al., Antimicrob Agents Chemother, July 2001; Bager et al., Microb Drug Resist, 1999).
• Decline in healthy people. In Saxony-Anhalt, VRE carriage in healthy non-hospitalised people fell from 12% (1994) to 3% (1997) after avoparcin was discontinued (Klare et al., Microb Drug Resist, 1999).
• Early transatlantic comparison. A 2001 comparison concluded that the European animal reservoir explained community VRE carriage in Europe. It also noted that avoparcin was never used in the US, where healthy people were not colonised (Bonten, Willems and Weinstein, Lancet Infect Dis, December 2001).
• WHO review. WHO’s independent review concluded that the Danish terminations had “dramatically reduced the food animal reservoir of enterococci resistant to these growth promoters”. It noted, however, that “clinical problems in humans related to resistance to antimicrobial growth promoters were rare in Denmark before and after termination” (WHO, Impacts of antimicrobial growth promoter termination in Denmark, WHO/CDS/CPE/ZFK/2003.1, 2003).

Evidence that clinical (hospital) VRE is largely a separate problem.

• The US case. The same 2001 review pointed out that the US hospital VRE epidemic arose with no avoparcin use. It attributed that epidemic to much heavier hospital vancomycin use and prior ampicillin resistance (Bonten et al., 2001).
• A distinct hospital complex. A study of 411 isolates from five continents identified a hospital-associated genetic complex (CC17) responsible for hospital outbreaks worldwide (Willems et al., Emerg Infect Dis, June 2005).
• Genome sequencing of the lineages.
• The epidemic hospital lineage emerged about 75 years ago, “concomitant with the introduction of antibiotics”, from a population that included most animal strains (Lebreton et al., mBio, 20 August 2013). This is a deep-time link, not evidence of avoparcin-era transmission.
• A UK One Health study compared more than 600 livestock, meat and wastewater isolates with nearly 800 bloodstream isolates. It found “the majority of human and livestock-related isolates were genetically distinct”, with “limited sharing of strains and resistance genes”. It found no VRE on 29 UK farms (Gouliouris et al., mBio, 6 November 2018).
• Reviews since 2012. These converge on a narrower claim. Animal E. faecium are mostly distinct from clinical strains but may act as “donors of antimicrobial resistance genes”, because the same Tn1546 variants appear in both. Animal E. faecalis may be a more direct hazard (Hammerum, Clin Microbiol Infect, April 2012; Bortolaia, Espinosa-Gongora and Guardabassi, Clin Microbiol Infect, 2016).
• The hospital VRE burden in Europe has grown since the ban. ECDC reported 16.5% vancomycin resistance in invasive E. faecium across the EU/EEA in 2024 (16.8% in 2020, so no EU-level trend over those five years):
• national values ranged from 0% to 61.7%, with ten countries above 25%;
• the estimated EU/EEA incidence of VRE E. faecium bloodstream infections was higher in 2024 than in 2020, though with no significant EU-level trend, and seven countries showed significantly increasing incidence trends;
• ECDC’s burden study estimated that VRE infections rose from 47,124 (2016) to 117,866 (2020), and attributable deaths from 1,335 to 3,414.

(ECDC, Antimicrobial resistance in the EU/EEA (EARS-Net), Annual Epidemiological Report for 2024, 18 November 2025). - Denmark, the first mover. Vancomycin resistance in invasive E. faecium has increased “for the last decade” and “surpassed 10% for the past four years”, reaching 13.9% in 2024. Vancomycin resistance in broiler enterococci was 0% (DANMAP 2024). - An early critic’s reading. Critics read the same pattern early. Casewell et al. (2003) attributed the rise in European human VRE infections “probably” to increased vancomycin use against MRSA, not to animals (J Antimicrob Chemother, July 2003).

Evidence on the general claim (animal use contributes to resistance in humans). This has been strengthened by later evidence, though mostly for resistance classes other than VRE.

• Colistin in China. Plasmid-mediated colistin resistance (mcr-1) was discovered in pigs in China in 2015 (Liu et al., Lancet Infect Dis, online 18 November 2015). After China banned colistin as a growth promoter on 30 April 2017, national surveillance recorded large falls:

(Wang et al., Lancet Infect Dis, online 4 June 2020; corroborated by Shen et al., Lancet Microbe, May 2020). This is the clearest natural experiment linking a growth-promoter ban to falling resistance in people. - Systematic reviews. See Claim 3.

Verdict: partly held up. The general claim that growth-promoter use contributes to resistance in humans has held up and has been strengthened, above all by colistin. Selection of VRE by avoparcin in animals, and its spread into healthy people’s gut flora, is confirmed. But the claim that VRE is the “most convincing” case of harm to human health has weakened:

• hospital VRE infections come mainly from hospital-adapted lineages;
• the US epidemic arose without avoparcin;
• European hospital VRE has risen long after the ban.

Implication for weight. The lesson that a feed additive can build a reservoir of resistance that reaches people carries strong weight. The specific VRE story should not be used as proof that the bans prevented clinical infections. It shows a reservoir that was removed, not a disease burden that was averted.

A second hindsight lesson is that the reservoir that gets regulated may not be the one driving the harm that is later measured (pp. 96–97).

Claim 2: The EU’s 1975 approval of tylosin and spiramycin “probably” drove macrolide resistance in pig enterococci and Campylobacter#

Original claim (p. 94). “Against the Swann recommendations, the EU accepted the macrolides tylosin and spiramycin as growth promoters in 1975. This has probably been one of the major reasons for the widespread macrolide resistance in, for example, enterococci and campylobacter from pigs.” (Table 9.2 garbles this entry as “tolysin and spiramycin still permitted as growth promoters as human equivalents”.)

Subsequent developments.

• Scale of use before the ban. Tylosin was Denmark’s main pig growth promoter. Use was 68,350 kg for growth promotion plus about 5,000 kg for therapy in 1996 (WHO 2003), and 62 tonnes in 1997 (DANMAP 2008).
• Enterococci. Erythromycin resistance in pig enterococci was “almost 90%” in 1995–97. After tylosin use fell in 1998–99 it dropped to 46.7% (E. faecium) and 28.1% (E. faecalis) by 2000. In broiler E. faecium it fell from 76.3% (1997) to 12.7% (2000) (Aarestrup et al., 2001).
• WHO’s panel reported that erythromycin resistance in pig E. faecium at slaughter fell from 80% (1997) to under 20% after the terminations. It judged that tylosin “as an antimicrobial growth promoter had a much greater impact on resistance to erythromycin … than tylosin use as a therapeutic agent” (WHO 2003).
• By 2008 erythromycin resistance in Danish pig E. faecium was 32% (DANMAP 2008).
• Campylobacter coli from pigs. Erythromycin resistance fell “from 68% in 1998 to 24% in 2003” after tylosin was withdrawn. It then plateaued, while therapeutic macrolide use increased (DANMAP 2008).
• DANMAP no longer routinely reports C. coli from pigs. In 2024, macrolides were still “often used to treat infections, particularly in pigs” (DANMAP 2024).
• Limits of the evidence.
• These declines show that growth-promoter use sustained high macrolide resistance. They cannot test the counterfactual of what would have happened without the 1975 EU approval.
• The data are almost entirely Danish.
• The plateau at 20–30% reflects continued therapeutic macrolide use.
• The WHO panel also noted “an indication” that erythromycin resistance in E. faecalis from Danish people rose after termination. It attributed this possibly to increased therapeutic tylosin use in pigs (WHO 2003).

Verdict: held up. It is strengthened for the underlying causal role of growth-promoter tylosin. The chapter’s hedge (“probably … one of the major reasons”) was appropriate. The surveillance after withdrawal is consistent with it and quantifies it.

Implication for weight. This is one of the better-evidenced mechanistic claims in the chapter. It can carry substantial weight as an example of a use-driven reservoir that shrank quickly once the use stopped.

The plateau adds a lesson the chapter lacks: residual use in a neighbouring category (here, therapy) sets a floor under the benefit (pp. 94–95).

Claim 3: Swann was “accurate” and “far-sighted”, its dilution rested on “narrow considerations of what was precisely known”, and Swann and the farmers have been “vindicated by history”#

Original claim (p. 98, with p. 94). The Swann Report rested on “a low level of scientific proof, but on a competent microbiological assessment”. Later research “confirm[s] that the Swann Report was both accurate in its evaluation of data at the time and far-sighted in its assessment of future trends”. The dilution was “based mainly on narrow considerations of what was precisely known rather than on taking account of what was not known”. Swann and the farmers “have been vindicated by history”.

Subsequent developments.

Mechanisms and policy: strongly vindicated.

• Horizontal transfer is established. Horizontal gene transfer across species and genera, which the chapter cites as the core of the vindication (p. 98), is now textbook microbiology. The 2013 Annex 3 update makes the same point (EEA 2013, p. 732).
• Policy converged on Swann’s principle.
• The EU deleted all antibiotic growth promoters except coccidiostats and histomonostats from 1 January 2006 (Regulation (EC) No 1831/2003, Article 11(2)).
• The US ended production uses of medically important drugs in 2017.
• WHO recommended complete restriction in 2017.
• The UN General Assembly acknowledged “the particular need to phase out the use of medically important antimicrobials for this purpose” in 2024. See Claim 8 for these developments.
• The US dilution story is a close parallel.
• After Swann, FDA set up a task force in 1970. In 1977 it proposed withdrawing penicillin and tetracycline for subtherapeutic feed use.
• It never held the hearings, and it formally withdrew the 1977 notices on 22 December 2011. Its reason was that it was pursuing “other ongoing regulatory strategies” and “would need to prioritize any withdrawal proceedings” (FDA, 76 FR 79697, 22 December 2011).
• Production claims for these drugs were not removed until 1 January 2017. That is 34 years of regulatory limbo, followed by a voluntary route (Claim 8).

Human-health consequences: supported in direction, not quantified.

• Tang et al. (2017). A systematic review and meta-analysis of 181 studies found that restricting antibiotic use in food animals reduced resistance in animals, “commonly” by 10–15 percentage points. In the 13 human studies meta-analysed, resistance was 24% lower in intervention groups, “with a stronger association seen for humans with direct contact with food-producing animals”. The authors cautioned: “The implications for the general human population are less clear” (Tang et al., Lancet Planet Health, November 2017).
• Scott et al. (2018). A rapid systematic review of 93 studies concluded that limiting animal use “probably reduces antimicrobial resistance in humans. The magnitude of the effect cannot be quantified” (Scott et al., Int J Antimicrob Agents, 2018).
• WHO (2017). WHO’s guideline recommended “complete restriction of use of all classes of medically important antimicrobials in food-producing animals for growth promotion”. It rated this a “strong recommendation, low quality evidence”. The strength was justified by the “potentially large human health benefits” and by the “relatively small or non-existent” downsides of restriction (WHO guidelines on use of medically important antimicrobials in food-producing animals, November 2017).
• O’Neill review (2015). The UK-commissioned Review on Antimicrobial Resistance found that of 139 academic studies, 100 (72%) found evidence of a link between antibiotic use in animals and resistance in humans, and seven (5%) argued there was none (O’Neill, Antimicrobials in agriculture and the environment, December 2015). This counts studies rather than weighing them.
• JIACRA IV (2024). The EU agencies’ joint analysis found that, for certain bacteria–drug combinations, resistance in people “was associated with AMR in bacteria from food-producing animals which, in turn, was related to AMC [antimicrobial consumption] in animals”. The strength of these associations “differed markedly” by drug class, organism and sector (ECDC/EFSA/EMA, JIACRA IV, EFSA Journal, 23 February 2024).
• Sceptical and nuanced readings persist in the mainstream literature. Chang et al. argued that “there are no data conclusively showing the magnitude of the threat”. They found direct infection from animal sources to be “small in comparison with the overall burden”, with gene transfer the hardest to study and possibly “the most important of all” (Chang et al., Evol Appl, 2015).

An unforeseen weakness in Swann’s own criterion. Swann allowed feed use of antibiotics with “little or no application as therapeutic agents in man or animals” (p. 94). Hindsight shows that this category moves over time.

• Vancomycin came into wide hospital use only from the mid-1970s, as the chapter itself notes (p. 94).
• Colistin was little used in human medicine for decades. It became a last-resort drug, and its feed use then produced mcr-1.
• Ionophores, which are “not used in humans”, were kept permitted as a class. They were later found to co-select for VRE (see Claim 4).

A classification based on current human importance can therefore be overtaken by later changes in medicine or by linked resistance.

Verdict: partly held up. Swann’s microbiological assessment and its policy logic have been vindicated, and the dilution story is echoed in the US record. But history’s verdict on the human-health consequences rests on evidence that WHO itself rates as low quality and that remains unquantified. The chapter’s “vindicated by history” was premature in 2001 on that dimension, and is still only partly earned in 2026.

Implication for weight. Treat this as strong support for the process lessons:

• acting on a competent mechanistic assessment rather than waiting for quantified harm;
• the hidden cost of letting a precautionary regime erode.

Do not treat it as evidence that the harms averted were large. Their size is still unknown. The chapter’s own lesson that “unknown magnitude is not unknown direction” (pp. 95–97) is the right way to carry it.

Claim 4: Sweden’s 1986 ban confined antimicrobials to veterinary prescription and, “as a result”, cut total use from about 50 to about 20 tonnes#

Original claim (p. 95). Since the 1986 Feedingstuffs Act, “antimicrobials, whether in feed or administered otherwise, have only been allowed for therapy and on veterinary prescription, and as a result the total consumption of antimicrobials was greatly reduced from around 50 tonnes in 1985 to around 20 tonnes in 1996 (SOU, 1997)”. No data are given on animal health or productivity.

Subsequent developments.

• Sweden’s official series. Sweden now publishes a population-corrected series (mg of active substance per population correction unit, PCU) from 1980 onward, split by growth promotion, group treatment and individual treatment. It omits 1985 and gives no total tonnages for the 1980s. Selected years, read from the interactive figure data in Swedres-Svarm 2025, Chapter 2, published 17 June 2026 (totals rounded):

So the ban removed growth promoters at once. But total use then climbed back by about 40% to 1992 as group medication rose, before falling in the mid-1990s. In mg/PCU, the 1984–1996 change is roughly a halving. That is consistent in direction with “50 to 20 tonnes”, but the path was not a simple result of the ban. - Swedres-Svarm’s own account. It attributes the long-run decline “first [to] the removal of growth-promoting antimicrobials in 1986, followed by a gradual but major decrease in the sales of veterinary products for group medication via feed or water from the mid-1990s onward”. Overall sales in 2025 were about 70% lower than the 1980–84 average (9,053 kg of active substance in 2025). - Animal health after the ban. The data the chapter omits were published the same year, in an account from within the Swedish veterinary authorities (Wierup, Microb Drug Resist, 2001). - Piglets. “Significant clinical problems emerged that created a demand for antibiotic-medicated feed at therapeutic dosages”. Use then “increased, involving up to 75% of the pigs”. It “could be halved in 1993” through better management, “supported by the addition of zinc oxide to the feed”. - Broilers. Expected necrotic enteritis “was prevented by a continuous use of antibiotics, largely to the same extent during the first 2 years after the ban”, until management changes made that unnecessary. - Other stock. Slaughter pigs, beef and turkeys showed “no negative clinical effects”. - Ionophore coccidiostats stayed permitted. Sweden’s prescription-only rule applied to antibacterial medicines. Ionophore coccidiostats remained feed additives. “Since the late 1980s, narasin has been by far the most widely used substance for broilers” (Swedres-Svarm 2025; also Swedres-Svarm 2024). - At EU level, coccidiostats were explicitly exempted from the 2006 phase-out (Regulation 1831/2003, Article 11). The Commission decided in 2008 not to phase them out, finding their preventive use “essential” in poultry (COM(2008) 233 final, 5 May 2008). - The ionophore exemption had a resistance cost that nobody anticipated in 2001. - Swedish broiler VRE isolates were found to carry transferable reduced susceptibility to narasin together with vanA (Nilsson et al., J Appl Microbiol, 2012). - Norway’s broiler industry stopped narasin in 2016. Within two years VRE fell “below the detection limit of the surveillance method” (Simm et al., PLoS One, December 2019). - The resistance mechanism, NarAB, was characterised in 2020 (Naemi et al., Front Microbiol, 2020). - A controlled broiler experiment published in August 2026 showed that “narasin use in broiler feed can co-select for vancomycin-resistant bacteria, including VREfm, through the NarAB mechanism” (Frederiksen et al., J Antimicrob Chemother, online 25 August 2026).

This helps explain why VRE persisted in Nordic broilers for two decades after avoparcin was banned. - Substitution continues. When the EU withdrew veterinary medicines with high levels of zinc oxide in 2022, Swedish sales of aminoglycosides for weaner diarrhoea rose. Neomycin resistance in pig E. coli then “necessitated the introduction of apramycin use in Sweden” (Swedres-Svarm 2025).

Verdict: partly held up. The long-run outcome strongly supports the Swedish approach: sustained low use, low resistance, and competitive production. But the specific “as a result” claim for 1985–96 compresses a non-monotonic path. It omits:

• the post-ban rise in therapeutic group medication;
• the role of zinc oxide and continued ionophore use;
• the real animal-health problems in weaners.

Implication for weight. Sweden is good evidence that production can adapt without routine antibiotic feeding. It is not evidence that a ban alone delivers the reduction, or that it does so smoothly.

Two technology-neutral lessons emerge that the chapter does not draw (pp. 95, 98):

1. Removing one input shifts dependence onto neighbouring inputs (therapy, zinc, ionophores). Each substitute can carry its own delayed cost.
2. An exempted category defined by “not used in human medicine” can still co-select for resistance that matters to human medicine.

Claim 5: Studies before the bans showed “wrong” the view that sub-inhibitory concentrations cannot select resistance#

Original claim (pp. 94–95). Walton (1988) argued that “the use of a sub-lethal or a sub-inhibitory antibiotic concentration is therefore unable to select resistant strains”, so that Swann was “in error”. “The recent bans on avoparcin, virginiamycin and tylosin followed the publication of studies demonstrating that this view (Walton’s amongst others) was wrong.”

Subsequent developments.

• What refuted Walton at the time were field studies. They showed selection at growth-promoting doses on farms (Bager et al., 1997; Aarestrup et al., 2001, and earlier Danish work). These show selection under real feed conditions. They do not establish whether the concentrations in the gut were below the inhibitory level for the relevant bacteria.
• Laboratory proof came later.
• Competition experiments showed that “drug concentrations up to several hundred-fold below the minimal inhibitory concentration of susceptible bacteria could enrich for resistant bacteria”. De novo mutants could also be selected below the MIC (Gullberg et al., PLoS Pathog, 21 July 2011).
• A follow-up showed that a large multidrug-resistance plasmid was maintained at “minimal selective concentrations” of antibiotics and heavy metals up to about 140-fold below the MIC. It also showed that combinations lowered these thresholds further (Gullberg et al., mBio, 7 October 2014).
• A review concluded that sub-inhibitory exposure is “likely to have an important role in the evolution of antibiotic resistance” in humans, livestock and the environment (Andersson and Hughes, Nat Rev Microbiol, 27 May 2014).
• Direct in-feed evidence. In a controlled pig experiment, a standard growth-promoting in-feed combination (chlortetracycline, sulfamethazine, penicillin) increased the abundance and diversity of resistance genes in the gut. This included genes conferring resistance to antibiotics “not administered in this study” (Looft et al., PNAS, 17 January 2012).
• A nuance. The later work also defines a minimal selective concentration, a threshold below which selection does not occur. So the refutation is not “any trace selects”. It is that the threshold lies far below the MIC, well within the range of growth-promoting doses.

Verdict: strengthened. Walton’s position has been decisively refuted by mechanistic evidence far stronger than what was available in 2001. The chapter’s evidential basis at the time was field association rather than proof at low concentrations, but its conclusion was correct.

Implication for weight. This is one of the chapter’s most robust claims. It is a clean example of the chapter’s “retreating reassurance” pattern (pp. 94–95, 98): a sincere model treated the edge of knowledge as a boundary of risk. The model was later overturned at the mechanistic level, not just the empirical one.

Claim 6: The virginiamycin manufacturer’s court challenge was pending, with no ruling expected “before the end of the year 2001”#

Original claim (p. 96; Table 9.2). The four-drug ban was “challenged before the European Court of Justice by the manufacturer of virginiamycin, who called for an annulment of the entire decision. Final ruling in the case is not expected before the end of the year 2001.”

Subsequent developments.

• The action and interim relief.
• Pfizer filed on 18 January 1999 before the Court of First Instance (now the General Court), not the Court of Justice.
• Its application to suspend the ban was dismissed on 30 June 1999 (T-13/99 R).
• Its appeal was dismissed by the President of the Court of Justice on 18 November 1999 (C-329/99 P(R)).
• The judgment. On 11 September 2002 the Court of First Instance (Third Chamber) “[d]ismisse[d] the application” and ordered Pfizer to pay the Council’s costs (Case T-13/99 Pfizer Animal Health v Council, ECLI:EU:T:2002:209). Alpharma’s parallel challenge to the bacitracin zinc ban was dismissed the same day (Case T-70/99 Alpharma v Council). I found no record of an appeal against either judgment.
• What the Court held. Several passages bear directly on the chapter’s lessons.
• Acting before harm is proven. “[W]here there is scientific uncertainty as to the existence or extent of risks to human health, the Community institutions may, by reason of the precautionary principle, take protective measures without having to wait until the reality and seriousness of those risks become fully apparent” (para. 139). A risk assessment “cannot be required to provide the Community institutions with conclusive scientific evidence of the reality of the risk” (para. 142).
• The limit on precaution. A preventive measure “cannot properly be based on a purely hypothetical approach to the risk, founded on mere conjecture which has not been scientifically verified”. It may be taken only if the risk “appears nevertheless to be adequately backed up by the scientific data available at the time” (paras 143–144). This qualifies the chapter’s framing: the Court endorsed precaution, but tied it to a scientific risk assessment.
• Departing from scientific advice. The Council had departed from its own scientific committee. In an opinion of 10 July 1998 that committee (SCAN) concluded that virginiamycin “does not constitute an immediate risk to public health in Denmark” and that a full risk assessment needed “quantitative evidence”. The Court held that an institution may disregard such an opinion if it gives “specific reasons … of a scientific level at least commensurate with that of the opinion” (para. 199). “Scientific legitimacy is not a sufficient basis for the exercise of public authority” (para. 201).
• Weighing the evidence. The Court found “great uncertainty” about the link to streptogramin resistance in humans. It relied on other bodies’ judgements to find that the Council “could properly find that there were serious reasons … for restricting streptogramins to medical use” (paras 393–402). Those bodies were the WHO 1997 meeting, the Copenhagen recommendations, the Swedish report (SOU 1997, the chapter authors’ own commission), the Netherlands Health Council and the House of Lords.
• Pfizer’s arguments. Pfizer argued that a “zero risk” test “would quickly lead to the paralysis of technological development and innovation” (para. 130). The Court rejected the claim that such a test had been applied (para. 403).
• Imports. Pfizer also argued that the ban was undermined because meat from animals fed the banned substances was still imported (para. 430). This is the gap the EU closed in September 2026 (Claim 8).
• Legal afterlife. EU Publications Office metadata record at least 98 later EU documents citing the judgment, including Advocate General opinions in the 2019–2020 pesticide cases (CELLAR citation query, 25 September 2026).

Verdict: held up. The chapter’s statement about timing was accurate. The case was dismissed in 2002, a year later than the chapter’s earliest date. The chapter misnamed the court.

Implication for weight. The outcome strengthens the case’s standing as a precedent for acting under uncertainty. But the reasoning adds nuance the chapter lacks:

• precaution was upheld because the institutions could point to a scientifically grounded risk and a documented reason for departing from their committee;
• it was not upheld because proof was unnecessary.

This supports the chapter’s committee-composition lesson (p. 98) in an unexpected way. The law allowed political decision-makers to overrule a narrowly constituted scientific committee, provided they gave reasons of comparable scientific quality.

Claim 7: Denmark’s 1998 virginiamycin ban aimed to “preserve the lifespan of Synercid”#

Original claim (p. 96). Denmark banned virginiamycin on 16 January 1998 “due to a risk of selection of streptogramin-resistant enterococci in pigs and poultry”, “to protect human health and to preserve the lifespan of Synercid, which was then undergoing hospital trials but which has now been licensed for the treatment of certain multi-drug-resistant infections in humans”.

Subsequent developments.

Resistance in animals and people.

• In animals. Virginiamycin resistance in Danish broiler E. faecium rose from 27.3% (1995) to 66.2% (1997) as use increased. After the ban it fell to 33.9% (2000) (Aarestrup et al., 2001). Resistance to quinupristin–dalfopristin (Q/D) in Danish broiler E. faecium was 2% in 2008 (DANMAP 2008).
• DANMAP 2024 reports 37% Q/D resistance in broiler E. faecium. It offers no comment, and panels and cut-offs have changed since 2008, so this should not be read as a rebound without checking the methods.
• In people. WHO’s panel found “some indication” of a decline in streptogramin resistance in human E. faecium in Denmark after the terminations (WHO 2003).
• In the US, where virginiamycin continued.
• Q/D-resistant E. faecium was found on 237 of 407 supermarket chickens (1998–99), but in only 3 of 334 outpatient stool samples, all low-level. The authors concluded virginiamycin “has not yet had a substantial influence” on people (McDonald et al., NEJM, 18 October 2001).
• A later study found poultry exposure associated with the resistance gene vatE and with inducible Q/D resistance in hospital patients’ gut E. faecium. vatE was present in 38% of patient isolates and none from vegetarians (Kieke et al., J Infect Dis, 2006).
• FDA published a draft quantitative risk assessment of streptogramin resistance attributable to animal use in November 2004 (69 FR 68384, 24 November 2004). I found no Federal Register notice of a final version.
• FDA classes streptogramins as medically important, so virginiamycin’s production claims ended with the 2017 changes (FDA sales data class lists; see Claim 8).

Synercid’s clinical course.

• FDA approved Synercid on 21 September 1999 (Drugs@FDA, NDA 050747/050748).
• Its VRE indication “was removed due to a failure to show a clinical benefit”. The drug has dose-limiting myalgias and arthralgias, needs central venous access, and “lacks activity against E. faecalis” (O’Driscoll and Crank, Infect Drug Resist, 24 July 2015).
• Linezolid, daptomycin and tigecycline displaced it as options for VRE.
• Drugs@FDA still lists one Synercid application as prescription status. But DailyMed held no current Synercid label when queried on 25 September 2026, which suggests it is little or not actively marketed in the US.
• SCAN’s judgment of the stakes has aged better than Denmark’s. In its 10 July 1998 opinion (quoted in T-13/99, para. 53), SCAN argued that Denmark had not detected “a single case of VRE” in its surveillance survey and had among the lowest MRSA rates. It concluded that “there are no clinical reasons to require the introduction of streptogramins as human therapeutics in Denmark now or in the immediate future”. In 1998 SCAN found transfer to the human gut unsubstantiated; later work (McDonald 2001; Kieke 2006) supports such a pathway. But its view of the drug’s clinical stakes turned out to be closer to what happened than the “lifespan of Synercid” rationale. The chapter mentions SCAN only for avoparcin (p. 96), and never mentions its dissent on virginiamycin.

Verdict: partly held up. The resistance dynamics the Danish ban targeted behaved as expected: a use-driven reservoir that shrank when use stopped. There is some evidence of a food-borne pathway to people’s gut flora. But the stake the ban was meant to protect proved smaller than expected, because Synercid turned out to be a marginal drug.

Implication for weight. Hindsight cuts both ways here.

• For precaution. At the time, Synercid was plausibly a last-resort option. The Court record describes it as potentially “the treatment of last resort … at least until other antibiotics … have been developed”. The decision was reasonable ex ante.
• Against. The benefit realised ex post was modest.

The technology-neutral lesson: preserving an option protects its expected value. Whether that value materialises depends on developments that cannot be known when the decision is made (p. 96).

Claim 8: The EU Scientific Steering Committee and WHO urged phase-out, while the US still used low-dose tetracycline and penicillin in feed without prescription#

Original claim (pp. 95, 97). The SSC (1999) recommended that growth promoters from classes “which are or may be used in human or veterinary medicine should be phased out as soon as possible and ultimately abolished”. WHO (2000) recommended that such use “should be terminated or rapidly phased out in the absence of risk-based evaluations”. At p. 95, “in some countries, such as the United States, low doses of tetracycline and penicillin are still used as feed additives for prophylaxis and growth promotion without veterinary prescription”.

Subsequent developments.

European Union.

• The 2006 phase-out. Regulation (EC) No 1831/2003 provides that “antibiotics, other than coccidiostats and histomonostats, may be marketed and used as feed additives only until 31 December 2005; as from 1 January 2006, those substances shall be deleted from the Register” (Article 11(2); Regulation (EC) No 1831/2003, 22 September 2003). The 1998 four-drug ban had applied from 1 January 1999, with a transition to 30 June 1999 in member states that had not already banned them (Council Regulation (EC) No 2821/98, 17 December 1998).
• The 2019 veterinary medicines regulation. Regulation (EU) 2019/6, applying from 28 January 2022, went beyond the growth-promoter ban (Regulation (EU) 2019/6). Article 107 provides:
• antimicrobials “shall not be applied routinely nor used to compensate for poor hygiene, inadequate animal husbandry or lack of care”;
• they “shall not be used in animals for the purpose of promoting growth nor to increase yield”;
• prophylaxis is allowed only in exceptional cases, and for antibiotics only in individual animals;
• metaphylaxis (group treatment when some animals are ill) is limited to cases of high risk with no alternatives.
• The ban extended to imports. Article 118 extends the growth-promotion ban to operators in third countries exporting to the EU. Delegated Regulation (EU) 2023/905 set the conditions (27 February 2023), which apply from 3 September 2026. From that date, animals and animal products may enter the EU only from listed third countries that have given guarantees. The list was first set by Implementing Regulation (EU) 2024/2598, 4 October 2024. Implementing Regulation (EU) 2026/1189, 4 June 2026 repealed that act and moved the list into Implementing Regulation (EU) 2021/405.
• The June 2026 act removed Brazil’s listing for bovine, equine, poultry, aquaculture, honey and casings. The Commission “has not received information guaranteeing” compliance by 3 September 2026 (recital 17).
• Brazil did not appear in Annex XVIa of the consolidated text applying from 3 September 2026 (consolidated Implementing Regulation (EU) 2021/405, version of 3 September 2026).

• News reports confirm the suspension came into force on 3 September 2026 while talks continued (secondary source). According to these reports:

  • Brazil had given written guarantees for poultry and honey but not for beef;
  • the Commission had begun an audit of Brazil’s poultry and honey production;
  • the Commission said beef needed guarantees covering the animals’ whole life cycle.

  (Euronews, 3 September 2026). As of this check I found no act relisting Brazil. - The first test of the trade extension. In its first month the import condition has shut out a major exporter for several product categories. The gap Pfizer pointed to in 1999 is being closed through market access, and that has trade consequences. The outcome of the audit, and whether Brazil is relisted, is still open.

This answers, 27 years later, Pfizer’s 1999 argument that the domestic ban was undermined by imports (T-13/99, para. 430). - Coccidiostats stay. The Commission reported in 2008 that it was “inappropriate to change the existing situation” (COM(2008) 233).

United States.

• Four decades of delay. FDA’s 1977 proposals to withdraw penicillin and tetracycline feed uses were never acted on, and were withdrawn in December 2011 (Claim 3).
• The voluntary route.
• FDA used voluntary guidance: GFI #209 (2012) and GFI #213 (2013), plus the Veterinary Feed Directive final rule (2015).
• It “[c]ompleted implementation of GFI #213” in January 2017, moving medically important antimicrobials in feed or water “from over-the-counter status to VFD or prescription status”.
• By June 2023, under GFI #263, all remaining sponsors had moved their other over-the-counter medically important products to prescription or withdrawn them.

(FDA, Timeline of FDA Action on Antimicrobial Resistance). - What happened to sales. FDA’s sales data (2015–2024 spreadsheet, FDA 2024 Summary Report, content current 21 November 2025; data file) show domestic sales of medically important antimicrobials for food animals:

(Totals are my sums from the file.) - Sales under “production” indications have been zero since 2017. - Tetracyclines were still 4,896 t (69%) of medically important sales in 2024, and 95% of medically important sales were for use in feed or water. - The chapter’s specific concern, low-dose tetracycline in feed without prescription, has therefore been addressed in legal form. In mass terms it persists as prescribed or VFD-authorised use for “prevention” and “control”. - Drugs not used in human medicine, mainly ionophores, still carried production claims: 4,384 t in 2024. - The WHO 2003 panel’s warning. It had anticipated this dynamic: “biology is not always changed by a legal definition, i.e., a law stating that they should not be used for prophylaxis does not stop the drugs from actually performing that function when they are used legally for another indication” (WHO 2003). - Other US restrictions. - FDA withdrew enrofloxacin for poultry in 2005 after a formal hearing (70 FR, 1 August 2005). - It prohibited certain extralabel cephalosporin uses from 2012 (77 FR 735, 6 January 2012).

Global.

• WHO (2017). WHO recommended complete restriction of growth-promoter use (Claim 3).
• WOAH (2025). WOAH’s ninth annual report (published 19 May 2025) found:
• 34 of 157 participating members (22%) still reported antimicrobial use for growth promotion, 80% of them in the Americas and Asia-Pacific;
• 112 (71%) reported no use;
• 11 did not know.

The molecules most often listed were bacitracin, tylosin and avilamycin, two of them growth promoters discussed in this chapter. About 7% of members reported using at least one highest-priority critically important antimicrobial, such as colistin, for growth promotion. No European member reported use (WOAH, Ninth Annual Report on Antimicrobial Agents Intended for Use in Animals, May 2025). The report is internally inconsistent about whether these data refer to 2022 or 2023. - China. China banned colistin as a growth promoter in April 2017 (Wang et al., 2020). Ministry of Agriculture and Rural Affairs Announcement No. 194 (2019) withdrew growth-promoting medicated feed additives, with implementation around 2020. The source for this is secondary: Hu and Cowling, Bull World Health Organ, 2020. I could not retrieve the Chinese ministry text. - UN General Assembly (2024). The political declaration on AMR (adopted 7 October 2024) “[a]cknowledge[s] the impact of antimicrobial growth promoters on antimicrobial resistance and the particular need to phase out the use of medically important antimicrobials for this purpose” (para. 66). It commits states to “[s]trive to meaningfully reduce, by 2030, the quantity of antimicrobials used globally in the agrifood system” (para. 69), but without a numerical target (A/RES/79/2).

Verdict: held up. The recommendations the chapter reported became policy in the EU (2006), the US (2017) and at the UN level (2024). The chapter’s description of US practice in 2001 was accurate.

Implication for weight. The direction of history clearly favours the chapter. The more useful hindsight lesson for the analytical lens is about implementation:

• The lags were long. The EU took 7 years after the SSC. The US took 17 years after WHO 2000 and 40 years after its own 1977 proposal. The world is still incomplete 25 years on.
• Purpose-based categories get relabelled. Once a use is banned by stated purpose, the same function can continue under a different stated purpose (“prevention” in place of “growth promotion”). This extends the chapter’s point about purpose-based classification (pp. 93, 95).
• Unilateral standards leak. A standard adopted by some jurisdictions is weakened by trade until it is extended to imports. That took 27 years here.

Claim 9: Human medical use “undoubtedly is of more importance”, but feed use is “one of the major sources of overuse and misuse” in animal farming#

Original claim (pp. 97–98). “Although the widespread use of antimicrobials in human medicine undoubtedly is of more importance for the emerging antimicrobial resistance problems in humans, this cannot justify ignorance of potential human health risks related to the use of antimicrobials in food animals. The continuous use of antimicrobials in feed is one of the major sources of overuse and misuse of antimicrobials in animal farming.”

Subsequent developments.

Scale of animal use.

• Global estimates. Global veterinary use was estimated at 93,309 t in 2017 (Tiseo et al., Antibiotics, December 2020) and 99,502 t in 2020. It is projected to reach 107,472 t (+8%) by 2030 (Mulchandani et al., PLOS Glob Public Health, February 2023). “Antimicrobial consumption in animals is threefold that of humans” (Van Boeckel et al., Science, 2019).
• EU/EEA. In 2021 consumption was 125.0 mg/kg biomass in humans and 92.6 mg/kg in food-producing animals. Animal consumption fell 44% from 2014 to 2021, while human consumption “remained relatively stable” (JIACRA IV, 2024). Animal use first fell below human use in the 2016–18 data (JIACRA III, EFSA Journal, June 2021).
• Denmark in 2024. Animals still used more by mass than humans: 87.46 t (73.40 t in pigs) against 51.57 t in human health care. Total animal use has fallen from the 1994 peak of 205.7 t (DANMAP 2024; WHO 2003).

Relative importance of human and animal sources for resistance in people.

• ESBL carriage in the Netherlands. A population-based model of community carriage of ESBL-producing E. coli attributed:
• 60.1% to human-to-human transmission;
• 18.9% to food;
• 7.9% to companion animals;
• 3.6% to farm animals through non-occupational contact.

(Mughini-Gras et al., Lancet Planet Health, August 2019). - UK genomic surveillance. Livestock and bloodstream-infection E. coli were “genetically distinct”, with “limited evidence that antimicrobial-resistant pathogens associated with serious human infection had originated from livestock” in the study region (Ludden et al., mBio, 22 January 2019). Similar findings for E. faecium appear in Gouliouris et al. (2018). - Cross-sector transfer can still be decisive for particular resistance genes. Colistin/mcr-1 is the standout case (Claim 1). The UN declaration now “[a]cknowledge[s] … the bidirectional spread of zoonotic disease and antimicrobial resistance between humans and animals” (A/RES/79/2, para. 68). - Scale of the overall problem. The UN declaration records 4.95 million deaths associated with bacterial AMR in 2019, and 1.27 million directly attributable to it. It sets a target to cut associated deaths by 10% by 2030. No consensus figure exists for the share attributable to animal use.

Verdict: held up. The chapter’s balanced position matches where the mainstream assessment now stands:

• human use is the dominant driver of resistance in human infections;
• animal use is large, often larger by mass, and a real contributor for specific organisms and genes;
• the size of animal use’s contribution to human disease burden remains unquantified.

Implication for weight. This is a well-calibrated claim and can carry substantial weight. It is a useful model of how to frame a secondary contributor: acknowledge the dominant source without letting it justify ignoring a large, controllable, secondary one (pp. 97–98).

Claim 10: Farmers’ voluntary withdrawal and “better animal husbandry” are viable alternatives#

Original claim (p. 98, with pp. 95–97 and Table 9.1). Farmers and their organisations “can take voluntary steps in advance of legislation”, and “they, as well as the Swann Committee, have been vindicated by history”. Risk assessments should consider “alternative options, such as better animal husbandry”. Table 9.1 lists no human-health benefit of growth promoters, and several ways they suppress alternatives.

Subsequent developments.

Evidence that the alternatives were viable.

• WHO international review panel (November 2002; published 2003). The panel concluded that “there have been no serious negative effects” and that “under conditions similar to those found in Denmark, the use of antimicrobials for the sole purpose of growth promotion can be discontinued”. Its specific findings:
• total food-animal antimicrobial use fell 54%, from 205,686 kg (1994) to 94,200 kg (2001);
• average exposure fell to 0.4 days per broiler and 7.9 days per pig;
• in broilers, feed efficiency fell 2.3% (0.9% when adjusted), “largely offset” by savings on growth promoters, with no change in weight gain or mortality;
• in pigs there was “some loss of productivity, primarily in weaners”, and no major effect in finishers;
• the net cost was about DKK 7.75 (€1.04) per pig, “just over 1%” of production costs, and an estimated 0.03% reduction in Danish GDP by 2010.

(WHO 2003). - Long-run pig data (Aarestrup et al., Am J Vet Res, July 2010). From 1992 to 2008: - antimicrobial use per kg of pig fell by more than half; - pigs per sow per year rose from 21 to 25; - average daily gain was higher in 2008 than in 1992; - mortality was similar.

The authors concluded that “long-term swine productivity was not negatively impacted”. - The US poultry case. The 2013 Annex 3 update reports the Perdue commercial trial, which covered some seven million birds. A reanalysis found that growth promoters produced a net loss of $0.0093 per chicken (Graham, Boland and Silbergeld, Public Health Rep, 2007; EEA 2013, pp. 732–733). The underlying trial was non-randomised.

Evidence of transitional costs and substitution that the chapter omits.

• The critics’ case. Casewell et al. (2003) argued that withdrawal “is now associated with a deterioration in animal health”. They cited:
• increased diarrhoea, weight loss and mortality from E. coli and Lawsonia intracellularis in weaned pigs;
• clostridial necrotic enteritis in broilers;
• more therapeutic use of drugs important in human medicine.

They concluded that the “theoretical and political benefit” of bans must be weighed against these costs (J Antimicrob Chemother, 2003). A broader critical review judged the actual danger to human health “small” and called precaution “a non-scientific approach” (Phillips et al., J Antimicrob Chemother, online 4 December 2003). Danish public-health scientists published a rebuttal questioning the review’s balance (Jensen et al., “An unbiased review?”, J Antimicrob Chemother, 2004). I could not access the full texts to check either side’s funding declarations. - What the WHO panel confirmed of the critics’ account. - Therapeutic treatment for post-weaning diarrhoea rose significantly, with increased use of tetracyclines, penicillins and macrolides. Total therapeutic use in 2000–01 was similar to the 1994 peak. - Weaner pigs showed reduced growth and increased mortality. - Necrotic enteritis in broilers stayed minor “largely because producers continued to use ionophores”. - Many weaner effects were “probably due to termination of olaquindox and carbadox”. These were withdrawn EU-wide in 1999 on grounds of occupational toxicity, and would have gone regardless. - Rebound in Danish pig use. Use per kg of pig rose from 31 mg (1999) to 49 mg (2008) (Aarestrup et al., 2010). It was brought down by further measures, including the Yellow Card herd-level thresholds from 2010 (DANMAP 2024). - DANMAP 2024 compares the 2022 zinc oxide withdrawal with the growth-promoter withdrawal: “antimicrobial treatment for diarrhoea doubled in the first year”, then “declined again as producers adapted”. - Sweden. Sweden had the same pattern of rebound, then adaptation, and relied on zinc oxide and ionophores along the way (Claim 4). - Limits on generalising the voluntary model. WHO’s panel noted that Denmark’s pig industry is “organized into a governing cooperative and farmers are co-owners of slaughterhouses”. “For this and perhaps other reasons, Denmark may be one of the few countries whose agricultural industry could implement a nation-wide voluntary action on antimicrobial growth promoters of this magnitude.” It added that countries with lower animal-health status might see different effects (WHO 2003). - Human health effects of the Danish withdrawal. Withdrawal “appears not to have affected” the incidence of human Salmonella, Campylobacter or Yersinia infections, or drug residues in food (WHO 2003). This is consistent with Table 9.1’s “None” for human-health benefits of growth promoters.

Verdict: partly held up. Viability is well supported: Danish and Swedish production grew, long-run productivity was maintained, and costs were small at national scale. But the alternatives were not purely “better husbandry”. They included:

• years of increased therapeutic use;
• zinc oxide, itself later withdrawn, with a new rebound;
• ionophores, later implicated in co-selecting for VRE.

The voluntary route depended on an unusual industry structure. The critics’ specific transitional observations were largely correct. Their prediction of lasting harm was not.

Implication for weight. The lesson that “it can be done” carries strong weight. The lesson that “voluntary action by producers leads the way” carries moderate weight, and should be conditioned on industry structure. Downstream users acted where their market depended on consumer trust and where they were organised collectively (pp. 95–96).

The chapter’s weakest element here is its silence on transition costs. Hindsight suggests a technology-neutral pattern: withdrawal of an embedded input produces an acute adjustment burden concentrated on the most vulnerable part of the system (here, weaners). Planning for that burden, as WHO’s panel recommended, is part of a credible precautionary policy (p. 98; Table 9.1).

What hindsight adds to the section’s lessons (technology-neutral)#

Each point below is tied to the page where the chapter makes, or should have made, the corresponding argument.

1. Linked effects defeat single-agent controls (pp. 94–96). - The chapter framed risk through cross-resistance pairs (pp. 94–97). - Hindsight shows co-selection: resistance to one agent can be maintained by a different agent whose resistance genes travel with it. Examples are pig glycopeptide resistance maintained by tylosin, and VRE in poultry maintained by narasin. - Controlling the named agent is not enough if linked agents stay in use.
2. Bans shift dependence onto neighbouring inputs (pp. 95, 98). Therapeutic group medication, zinc oxide and ionophores each absorbed part of the function growth promoters had served. Some later carried their own delayed costs.
3. Regulating by stated purpose invites relabelling (pp. 93, 95). - The chapter already saw growth promoters working as disguised prophylaxis (p. 95, SOU 1997). - After purpose-based bans, the same drugs continued in large quantities under “prevention” and “control” indications, most visibly in the US after 2017. WHO’s 2003 panel named this directly.
4. “Low current importance” is a moving target (p. 94). Swann’s criterion of “little or no application as therapeutic agents” was sound when written. It was overtaken when vancomycin and later colistin became critical, and when a “non-human” class (ionophores) turned out to co-select for resistance to a human drug.
5. Monitoring built after action produces the evidence that later judges the action (pp. 94, 97). Almost everything that “vindicates” the bans comes from DANMAP and Swedres-Svarm. Neither existed when Swann recommended monitoring in 1969. Where monitoring was not built, as in the US, the debate stayed open longer.
6. The law’s version of precaution is conditional (pp. 96, 98). Courts upheld action under uncertainty, but required a scientifically grounded, non-hypothetical risk and reasoned departure from expert advice. This is narrower and more demanding than the chapter’s “common sense” framing.
7. Unilateral standards leak until extended to trade (pp. 95–96). The EU’s import condition, applying from 3 September 2026, closes a gap the manufacturer pointed to in 1999. Its first application shut out a major exporter that could not give guarantees, which shows that closing the leak has trade costs of its own.
8. The regulated reservoir may not be the one driving the harm later measured (pp. 96–97). Removing the animal VRE reservoir did not prevent a later hospital VRE rise driven largely by hospital lineages. Precautionary action can be justified and successful on its own terms without being the solution to the downstream problem people later associate with it.

Sources#

Primary: law, regulation and court documents - Council Regulation (EC) No 2821/98 of 17 December 1998 (withdrawal of bacitracin zinc, spiramycin, virginiamycin, tylosin phosphate). OJ L 351, 29.12.1998. https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:31998R2821 - Case T-13/99 Pfizer Animal Health SA v Council, Judgment of the Court of First Instance (Third Chamber), 11 September 2002, ECLI:EU:T:2002:209. https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:61999TJ0013 - Case T-70/99 Alpharma Inc. v Council, Judgment of 11 September 2002. https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:61999TJ0070 - Case C-329/99 P(R) Pfizer Animal Health v Council, Order of the President of the Court of Justice, 18 November 1999, ECLI:EU:C:1999:572. https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:61999CO0329 - Regulation (EC) No 1831/2003 on additives for use in animal nutrition, 22 September 2003 (Art. 11). https://eur-lex.europa.eu/eli/reg/2003/1831/oj - European Commission, COM(2008) 233 final, Report on the use of coccidiostats and histomonostats as feed additives, 5 May 2008. https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:52008DC0233 - Regulation (EU) 2019/6 on veterinary medicinal products, 11 December 2018, applicable from 28 January 2022 (Arts 107, 118). https://eur-lex.europa.eu/eli/reg/2019/6/oj - Commission Delegated Regulation (EU) 2023/905, 27 February 2023. https://eur-lex.europa.eu/eli/reg_del/2023/905/oj - Commission Implementing Regulation (EU) 2024/2598, 4 October 2024 (first list of third countries; repealed by Implementing Regulation (EU) 2026/1189). https://eur-lex.europa.eu/eli/reg_impl/2024/2598/oj - Commission Implementing Regulation (EU) 2026/1189, 4 June 2026, amending Implementing Regulation (EU) 2021/405 and repealing Implementing Regulation (EU) 2024/2598 (Art. 1 applies from 3 September 2026; recital 17 on Brazil). https://eur-lex.europa.eu/eli/reg_impl/2026/1189/oj - Commission Implementing Regulation (EU) 2021/405, consolidated version of 3 September 2026 (Annex XVIa). https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:02021R0405-20260903 - EU Publications Office CELLAR SPARQL endpoint (citations of 61999TJ0013; acts referencing 32023R0905), queried 25 September 2026. https://publications.europa.eu/webapi/rdf/sparql - US FDA, “Withdrawal of Notices of Opportunity for a Hearing; Penicillin and Tetracycline Used in Animal Feed”, 76 FR 79697, 22 December 2011. https://www.federalregister.gov/documents/2011/12/22/2011-32775/withdrawal-of-notices-of-opportunity-for-a-hearing-penicillin-and-tetracycline-used-in-animal-feed - US FDA, “Draft Risk Assessment of Streptogramin Resistance in Enterococcus faecium Attributable to the Use of Streptogramins in Animals; Availability”, 69 FR 68384, 24 November 2004. https://www.federalregister.gov/documents/2004/11/24/04-25979/draft-risk-assessment-of-streptogramin-resistance-in-enterococcus-faecium-attributable-to-the-use-of - US FDA, Enrofloxacin for poultry; withdrawal of approval, 1 August 2005. https://www.federalregister.gov/documents/2005/08/01/05-15223/animal-drugs-feeds-and-related-products-enrofloxacin-for-poultry-withdrawal-of-approval-of-new - US FDA, Cephalosporin drugs; extralabel use; order of prohibition, 6 January 2012. https://www.federalregister.gov/documents/2012/01/06/2012-35/new-animal-drugs-cephalosporin-drugs-extralabel-animal-drug-use-order-of-prohibition - US FDA, Timeline of FDA Action on Antimicrobial Resistance (accessed 25 September 2026). https://www.fda.gov/animal-veterinary/antimicrobial-resistance/timeline-fda-action-antimicrobial-resistance - US FDA, 2024 Summary Report on Antimicrobials Sold or Distributed for Use in Food-Producing Animals (content current 21 November 2025) and data file “CVM Antimicrobial Sales and Distribution Data 2015–2024”. https://www.fda.gov/animal-veterinary/antimicrobial-resistance/2024-summary-report-antimicrobials-sold-or-distributed-use-food-producing-animals; https://www.fda.gov/media/189753/download?attachment - Drugs@FDA via openFDA, Synercid (NDA 050747, 050748), queried 25 September 2026. https://api.fda.gov/drug/drugsfda.json?search=products.brand_name:SYNERCID - DailyMed SPL search for Synercid (0 results), queried 25 September 2026. https://dailymed.nlm.nih.gov/dailymed/services/v2/spls.json?drug_name=synercid - UN General Assembly resolution A/RES/79/2, Political declaration of the high-level meeting on antimicrobial resistance, adopted 7 October 2024. https://documents.un.org/api/symbol/access?s=A/RES/79/2&l=en&t=pdf

Primary: international and national reports and surveillance - WHO, Impacts of antimicrobial growth promoter termination in Denmark: the WHO international review panel’s evaluation, Foulum, 6–9 November 2002, WHO/CDS/CPE/ZFK/2003.1 (2003). https://iris.who.int/handle/10665/68357 - WHO, WHO guidelines on use of medically important antimicrobials in food-producing animals, November 2017. https://iris.who.int/handle/10665/258970 (executive summary: https://iris.who.int/handle/10665/259240) - WOAH, Ninth Annual Report on Antimicrobial Agents Intended for Use in Animals, 19 May 2025. https://www.woah.org/app/uploads/2025/05/en-woah-amu-report-2025-final.pdf - ECDC, Antimicrobial resistance in the EU/EEA (EARS-Net), Annual Epidemiological Report for 2024, 18 November 2025. https://www.ecdc.europa.eu/sites/default/files/documents/antimicrobial-resistance-eu-annual-epidemiological-report-2024.pdf - ECDC/EFSA/EMA, JIACRA IV (2019–2021), EFSA Journal 22:e8589, 23 February 2024. https://doi.org/10.2903/j.efsa.2024.8589 - ECDC/EFSA/EMA, JIACRA III (2016–2018), EFSA Journal 19:e06712, June 2021. https://doi.org/10.2903/j.efsa.2021.6712 - DANMAP 2008 (Denmark). https://www.danmap.org/-/media/institutter/foedevareinstituttet/publikationer/pub-2008/danmap_2008.pdf - DANMAP 2024 (Denmark), full report, 2025. https://www.danmap.org/reports/2024 - Swedres-Svarm 2016 (Sweden). https://www.sva.se/media/dpcfdcr1/rapport_swedres-svarm_2016.pdf - Swedres-Svarm 2024 (Sweden). https://www.sva.se/media/amupibfr/swedres-svarm-2024-webb.pdf - Swedres-Svarm 2025 (Sweden), published 17 June 2026; Chapter 2, Sales of antibiotics in animals. https://folkhalsomyndigheten-rapporter.github.io/swedres-svarm-2025/chapters/2_Sales_antibiotics_animals/Main.html - O’Neill J (Review on Antimicrobial Resistance), Antimicrobials in agriculture and the environment: reducing unnecessary use and waste, December 2015. https://amr-review.org/sites/default/files/Antimicrobials%20in%20agriculture%20and%20the%20environment%20-%20Reducing%20unnecessary%20use%20and%20waste.pdf - Euronews, “EU suspension of Brazilian meat comes into force, despite ongoing talks with Brasília”, 3 September 2026 (news; secondary source). https://www.euronews.com/2026/09/03/eu-suspension-of-brazilian-meat-comes-into-force-despite-ongoing-talks-with-brasilia - EEA, Late lessons from early warnings: science, precaution, innovation, EEA Report No 1/2013, Annex 3, “Antimicrobials as growth promoters” (E. K. Silbergeld), pp. 731–734; Annex 2, Table A2.9. https://www.eea.europa.eu/en/analysis/publications/late-lessons-2/late-lessons-2-full-report/late-lessons-2-annexes

Peer-reviewed studies and reviews - Aarestrup FM et al. (2001) Antimicrob Agents Chemother 45:2054–2059, July 2001. https://doi.org/10.1128/AAC.45.7.2054-2059.2001 - Aarestrup FM et al. (2010) Am J Vet Res 71:726–733, July 2010. https://doi.org/10.2460/ajvr.71.7.726 - Andersson DI, Hughes D (2014) Nat Rev Microbiol 12:465–478, 27 May 2014. https://doi.org/10.1038/nrmicro3270 - Bager F et al. (1997) Prev Vet Med 31:95–112, July 1997. https://doi.org/10.1016/S0167-5877(96)01119-1 - Bager F et al. (1999) Microb Drug Resist 5:53–56. https://doi.org/10.1089/mdr.1999.5.53 - Bonten MJ, Willems R, Weinstein RA (2001) Lancet Infect Dis 1:314–325, December 2001. https://doi.org/10.1016/S1473-3099(01)00145-1 - Bortolaia V, Espinosa-Gongora C, Guardabassi L (2016) Clin Microbiol Infect 22:130–140. https://doi.org/10.1016/j.cmi.2015.12.003 - Casewell M et al. (2003) J Antimicrob Chemother 52:159–161, July 2003. https://doi.org/10.1093/jac/dkg313 - Chang Q et al. (2015) Evol Appl 8:240–247. https://doi.org/10.1111/eva.12185 - Frederiksen RF et al. (2026) “Narasin used as a feed additive in conventional rearing of broilers can co-select for vancomycin-resistant Enterococcus faecium through the NarAB ionophore resistance mechanisms”, J Antimicrob Chemother 81(9):dkag253, online 25 August 2026. https://doi.org/10.1093/jac/dkag253 - Gouliouris T et al. (2018) mBio 9:e01780-18, 6 November 2018. https://doi.org/10.1128/mBio.01780-18 - Graham JP, Boland JJ, Silbergeld E (2007) Public Health Rep 122:79–87. https://doi.org/10.1177/003335490712200111 - Gullberg E et al. (2011) PLoS Pathog 7:e1002158, 21 July 2011. https://doi.org/10.1371/journal.ppat.1002158 - Gullberg E et al. (2014) mBio 5:e01918-14, 7 October 2014. https://doi.org/10.1128/mBio.01918-14 - Hammerum AM (2012) Clin Microbiol Infect 18:619–625. https://doi.org/10.1111/j.1469-0691.2012.03829.x - Hu YJ, Cowling BJ (2020) Bull World Health Organ 98:360–361 (secondary source for China’s 2019 Announcement No. 194). https://doi.org/10.2471/BLT.19.243501 - Jensen VF et al. (2004) “Does the use of antibiotics in food animals pose a risk to human health? An unbiased review?” J Antimicrob Chemother 54:274–275. https://doi.org/10.1093/jac/dkh264 - Kieke AL et al. (2006) J Infect Dis 194:1200–1208. https://doi.org/10.1086/508189 - Klare I et al. (1999) Microb Drug Resist 5:45–52. https://doi.org/10.1089/mdr.1999.5.45 - Lebreton F et al. (2013) mBio 4:e00534-13, 20 August 2013. https://doi.org/10.1128/mBio.00534-13 - Liu YY et al. (2016) Lancet Infect Dis 16:161–168, online 18 November 2015. https://doi.org/10.1016/S1473-3099(15)00424-7 - Looft T et al. (2012) PNAS 109:1691–1696, 17 January 2012. https://doi.org/10.1073/pnas.1120238109 - Ludden C et al. (2019) mBio 10:e02693-18, 22 January 2019. https://doi.org/10.1128/mBio.02693-18 - McDonald LC et al. (2001) N Engl J Med 345:1155–1160, 18 October 2001. https://doi.org/10.1056/NEJMoa010805 - Mughini-Gras L et al. (2019) Lancet Planet Health 3:e357–e369, August 2019. https://doi.org/10.1016/S2542-5196(19)30130-5 - Mulchandani R et al. (2023) PLOS Glob Public Health 3:e0001305, February 2023. https://doi.org/10.1371/journal.pgph.0001305 - Naemi AO et al. (2020) Front Microbiol 11:104. https://doi.org/10.3389/fmicb.2020.00104 - Nilsson O et al. (2012) J Appl Microbiol 112:716–722. https://doi.org/10.1111/j.1365-2672.2012.05254.x - O’Driscoll T, Crank CW (2015) Infect Drug Resist 8:217–230, 24 July 2015. https://doi.org/10.2147/IDR.S54125 - Phillips I et al. (2004) J Antimicrob Chemother 53:28–52, online 4 December 2003. https://doi.org/10.1093/jac/dkg483 - Scott AM et al. (2018) Int J Antimicrob Agents 52:316–323. https://doi.org/10.1016/j.ijantimicag.2018.04.005 - Shen C et al. (2020) Lancet Microbe 1:e34–e43, May 2020. https://doi.org/10.1016/S2666-5247(20)30005-7 - Simm R et al. (2019) PLoS One 14:e0226101, December 2019. https://doi.org/10.1371/journal.pone.0226101 - Tang KL et al. (2017) Lancet Planet Health 1:e316–e327, November 2017. https://doi.org/10.1016/S2542-5196(17)30141-9 - Tiseo K et al. (2020) Antibiotics 9:918, December 2020. https://doi.org/10.3390/antibiotics9120918 - Van Boeckel TP et al. (2019) Science 365:eaaw1944. https://doi.org/10.1126/science.aaw1944 - Wang Y et al. (2020) Lancet Infect Dis 20:1161–1171, online 4 June 2020. https://doi.org/10.1016/S1473-3099(20)30149-3 - Wierup M (2001) Microb Drug Resist 7:183–190. https://doi.org/10.1089/10766290152045066 - Willems RJ et al. (2005) Emerg Infect Dis 11:821–828, June 2005. https://doi.org/10.3201/eid1106.041204