LL2-A3 — Annex 3 An update of some case studies from vol. 1#

Report: Late lessons from early warnings: science, precaution, innovation (EEA Report No 1/2013), Annex 3: “An update of some case studies from Late lessons from early warnings: the precautionary principle 1896–2000 vol. 1” Report pages: 717–737. PDF pages: 719–739.

How this was read. I read the full text extract in order, every page, through to the final marker (PDF 739 / p. 737). I also rendered pp. 717, 718, 724, 725, 728, 731, 733, 735, 736 and 737 from the PDF and checked them visually. The extraction is faithful. The annex has no figures, tables or boxes, one footnote (p. 736) and several indented block quotations (pp. 724–725). About a quarter of the extract (by characters) is reference lists.

Conventions. Page citations are to the printed report page. Anything marked [External — verify] comes from my background knowledge, not from the annex. It flags contestation or later developments for the hindsight pass and has not been checked against sources in this step.

Dating. The annex cites events up to 13 March 2013 (European Parliament vote, p. 724) and a web page accessed on 7 March 2013 (p. 719). It describes the Minamata mercury treaty as “will be signed … in October 2013” (p. 724). So it was finalised around March–May 2013, which fits the “05-2013” date in the PDF’s filename. [External — verify: the main report was launched in January 2013, so this annex may have been finalised or revised after the launch.]

Authors and standpoint#

What kind of section this is#

Annex 3 is not a case study. It is a collection of nine short updates to nine Volume 1 (2001) case studies, each written separately, under an unsigned editorial introduction (p. 717). There are no panels, no industry responses, and no dissenting commentaries. The only institutional voice is the European Commission’s lawyer, writing about a dispute the Commission was party to. The pieces differ sharply in genre: scientific updates, a declared advocacy campaign, an anonymous summary of an intergovernmental science assessment, and a legal summary written by a lawyer for one party to a trade dispute.

Standpoints in more detail#

• Editors (p. 717). They present the updates as evidence for a central Vol. 2 thesis, that harm “expands over time”. They also openly endorse Ruzzin’s contribution as an early warning “in the spirit of the Late Lessons project”, adding that it has been “greeted in some quarters an ‘inconvenient truth’” (p. 717). The editorial stance is pro-precaution and treats the updates as confirming evidence.
• Who the authors are. Nearly all were already on record on one side. Four are original chapter authors revisiting their own cases (Infante, Gee, Koppe, Swan). Three are researchers whose own laboratory’s results form the core evidence (Ruzzin, Belpoggi, and Silbergeld’s reanalysis). One is the European Commission’s lawyer summarising a case the Commission litigated (Christoforou). No contributor comes from industry, an agriculture or trade ministry, or any body that disputed the precautionary position. No piece carries a declaration of interests. That does not make the content wrong. It does mean the annex is best read as a set of informed advocacy positions and status reports, not as a balanced review.
• Infante. His affiliation is a private consultancy. The reference list shows he was lead author of the 1977 Lancet paper “Leukaemia in benzene workers” (Infante et al., 1977, p. 722), which heads the list of Pliofilm-cohort publications cited on p. 721, so he is re-examining a cohort he helped describe. [External — verify: Infante previously worked at NIOSH and OSHA. He has also acted as an expert witness in benzene/gasoline litigation, and at least one US federal court (c. 2015) excluded his opinion that gasoline causes AML. This matters for his claim that gasoline “should be considered a cause of all” benzene-associated blood cancers (p. 720).]
• Castleman and Gee. No affiliations are printed. [External: David Gee was the EEA’s lead on the Late Lessons project. Barry Castleman is the author of the standard medico-legal history of asbestos, a long-time campaigner for a global ban, and has served as an expert in asbestos litigation. Verify details.]
• Belpoggi. She directs the laboratory whose rodent bioassays are the (uncited) basis of her MTBE claims. [External — verify: some Ramazzini Institute bioassays, including on methanol and MTBE, were questioned around 2010–2011 in a US pathology review of their lymphoma/leukaemia diagnoses. The US EPA paused reliance on some Ramazzini data. This bears directly on her claims about lymphomas and leukaemias.]
• Christoforou. Counsel for the EC. He calls the piece a “personal summary” (p. 736), but it reads as one party’s account of the litigation.
• Ozone summary. It has no human author voice. It simply summarises an intergovernmental assessment that was already about six years old when the annex was written (see below).

Section-by-section notes#

0. Editorial introduction (p. 717)#

• Rationale. “Important developments” in several Vol. 1 case studies merited “a brief update”. Asking Vol. 1 authors for comprehensive updates “was a step too far”, so the annex contains “short updates focusing on the most important developments … since 2001” (p. 717).
• Replacement authors. Some original authors “had subsequently died or could not be contacted”. Where issues were “of increasing importance” or there had been “a very significant legal development”, the editors “found other experts” (p. 717). This explains the Silbergeld, Belpoggi and Christoforou pieces and the anonymous ozone summary.
• Fisheries framed as a new warning. The editors describe it as “a different and new early warning, about the health benefits of farmed fish”. It is not an update of MacGarvin’s chapter, which was about fish stocks (p. 717).
• Headline synthesis. The updates “illustrate the nature and extent of continuing damage to ecosystems and people since Volume 1” (p. 717). The example given is breast cancer “now arising in the daughters of the mothers who took the pregnancy pill, DES, and who survived the vaginal cancer of their twenties” (p. 717). The lesson drawn: “for most of the harmful agents described in the case studies, their capacity to harm expands over time” to “other kinds of harm”, often at exposures “previously considered ‘safe’” (p. 717).
• Analytical comments.
• The DES example is loosely phrased and compresses two separate findings. Swan’s update reports raised breast cancer risk after age 40 among DES-exposed daughters as a group (p. 729). The raised risk of clear-cell adenocarcinoma is a separate finding. The intro’s wording (“who survived the vaginal cancer of their twenties”) can be read as saying the breast cancers occur specifically in survivors of that vaginal cancer; the update does not say so. (It may simply mean the cohort had passed through the age of highest vaginal-cancer risk.)
• The claim covers “most” agents, but the annex’s support is uneven. Benzene and DES support it well. Asbestos (school exposures, lower recommended limits) supports it in part. PCBs and fish support it weakly: PCBs largely by assertion, fish by the author’s two rodent studies plus mixed human associations. The ozone update is a recovery story (p. 728) and the hormones update is purely legal (pp. 736–737). Neither illustrates “expanding harm”, and the editors don’t say so.
• Only nine of the fourteen Vol. 1 case studies are updated. [External — verify: the ones not updated appear to be radiation (X-rays), sulphur dioxide/acid rain, Great Lakes contamination, TBT antifoulants and BSE.] The annex does not explain the choice beyond availability and importance.

1. Fisheries/aquaculture: emerging challenges — Ruzzin (pp. 718–719)#

Context (p. 718). Aquaculture supplies “more than half” of fish and shellfish eaten by humans, and its pressure on forage fisheries is “a source of intense debate” (Naylor et al., 2009). The health benefits of fatty fish and long-chain omega-3s were researched decades ago, and cardioprotection was reported (Kromhout et al., 1985).

The anomaly (p. 718). More recent studies found fish consumption “could induce a higher risk for type 2 diabetes” (Zhou et al., 2012, a meta-analysis). In one study, four servings a week went with a “49 % increased risk” (Djoussé et al., 2011). The cardiac benefits of omega-3 “have also seriously been questioned” (Rizos et al., 2012; Saravanan et al., 2010), and “a possible link” between omega-3 and prostate cancer “has even been reported” (Dahm et al., 2012). Ruzzin proposes two explanations: environmental pollutants and “the new nutritional value of fish”.

Environmental pollutants (p. 718). - In 2010, rats fed fish oil from farmed Atlantic salmon developed “multiple disorders associated with type 2 diabetes as well as obesity”. When the same oil was purified of POPs, the animals were protected (Ruzzin et al., 2010). - Mice fed commercially available farmed salmon fillet developed “severe metabolic dysfunctions” mimicking diabetes, obesity and fatty liver disease. Fillet with less POP content gave a better metabolic profile (Ibrahim et al., 2011). - His inference: “background levels of POPs, which many people consider to be at safe levels, can fully abolish the potential health benefits of fish nutrients” (p. 718). This fits “the emerging view that POPs are contributors to the epidemic of type 2 diabetes” (Alonso-Magdalena et al., 2011; Neel and Sargis, 2011; UNEP and WHO, 2013).

Nutritional value (p. 718). - Because of rapid aquaculture growth and limited marine resources, fishmeal and fish oil in aquafeeds “have been largely replaced by vegetable oil and protein”. - “Important financial resources” went into selecting vegetable ingredients “to maximize the growth and health of farmed fish”. - As a result, the ratio of EPA and DHA to n-6 PUFAs in farmed Atlantic salmon fell “by about 50 %” between 2005 and 2009 (NIFES data). - The impact of this fall “has not been researched, but it is likely to have significally reduced the health benefits of farmed fish” (sic).

Conclusion and recommendations (p. 718). - There is “‘early warnings’ evidence that the health benefits of seafood and fish may be over‑estimated”. - Pollutants in marine food in the EU are “still poorly regulated compared with other dietary products”. Organochlorine pesticides, some PCBs, PBDEs and perfluorinated compounds are “still unregulated in seafood intended for human consumption” (Ruzzin, 2012). - The common use of vegetable feed ingredients “introduces new hazardous chemicals that may threaten human health” (not named, not cited). - There is an “urgent need” to control pollutant levels in seafood and to document the health impacts of current farmed fish. Ignoring this “may have important public health consequences”.

Appraisal. - The causal core is two rodent feeding studies from the author’s own group. Three of the 14 references are his. - The human evidence is associational and heterogeneous. The nutritional-change claim is openly unresearched. The “new hazardous chemicals” are unnamed. - It is structurally a textbook early warning: a plausible mechanism, animal data, a regulatory gap and population-scale exposure. It is not demonstrated harm. - There is an analytically interesting point: the feed system was optimised for the growth and health of the fish (producer-relevant) and not for the nutritional value of the fish to humans (consumer-relevant). The consumer-relevant property drifted. Its composition was measured (the 50% figure comes from Norway’s national nutrition and seafood institute, NIFES), but the health consequences of the change were, on Ruzzin’s account, “not researched”. - The piece does not weigh benefits against risks. - [External — verify: a 2010 FAO/WHO expert consultation concluded that for most populations the benefits of eating fish outweigh the risks from methylmercury and dioxins. EU maximum levels for six indicator non-dioxin-like PCBs (Reg. (EU) 1259/2011, applying from 2012) already existed when the annex was written, so Ruzzin’s “some PCBs” refers to congeners outside that set. Later, EU maximum levels for certain PFAS in fish (Reg. (EU) 2022/2388, applying from 2023) partly closed the gap he describes. The later evidence on fish, omega-3, diabetes and prostate cancer is mixed and population-dependent.]

2. Benzene and gasoline — Infante (pp. 720–723)#

Sources (p. 720). Infante opens with the unsourced statement that benzene “is the most toxic aromatic hydrocarbon”. Exposure comes from the oil, chemical and paint industries, vehicle exhaust, gasoline vapour, and household sources (solvent use, paints, leaking underground tanks).

History and downward drift of effect levels (p. 720). Benzene has been known since 1897 (LeNoir and Claude) to cause fatal blood diseases at high exposure. Research since then has shown effects on the blood-forming system “from lower and lower benzene exposure levels”, now down to average exposures “below 1 ppm”.

Cancer spectrum claimed (p. 720). - He claims significantly elevated risk of “all of the major forms of leukemia”: AML, ALL, CML, NHL, multiple myeloma and CLL (Infante and Bingham, 2012; Infante, 2006; Glass, 2003). - Recent studies show elevated AML and MM at average exposures below 1 ppm (Kirkeleit et al., 2008; Talbott et al., 2011). - Supporting evidence: lower white cell counts and damage to blood stem cells and leukocytes among workers exposed “down to the lowest levels evaluated (0.1–0.2 ppm)” (Lan et al., 2004, 2006; Qu et al., 2002). - For NHL: experimental lymphoma induction plus mechanistic knowledge give “a high level of biomedical plausibility” for a causal link with MM, CLL and lymphomas (Infante, 2011; Goldstein, 2009).

Gasoline (p. 720). - Gasoline is associated with elevated AML and cytopenias from workplace and residential exposures. - These are the most commonly observed blood diseases among benzene-exposed people, which Infante says “may be a reflection of” higher background rates, shorter latency and a more potent effect. - Because gasoline contains “at least 1–2 %” benzene today, it “should be considered a cause of all lymphohematopoietic cancers associated with benzene exposure to both adults and children” (Infante and Bingham, 2012). - Note: this is a causal attribution inferred from benzene content. It does not rest on gasoline-specific epidemiology for each cancer type.

Childhood leukaemia (p. 720). - ALL and AML “appear” significantly elevated with exposures before birth and in childhood from gasoline, paints and second-hand cigarette smoke. These exposures were “likely” well below 1 ppm. - Evidence cited: mothers exposed at work (Shu et al., 1988); painting at home during pregnancy (Shu et al., 1999; Schüz et al., 2000; Freedman et al., 2001; Bailey et al., 2011); living near petrol stations (Steffen et al., 2004; Brosselin et al., 2009). - Animal support: exposure before birth induces blood-system tumours in offspring (Maltoni et al., 1989; Badham et al., 2010a) and oxidative stress in embryonic signalling (Badham et al., 2010b).

Quantitative risk assessment (QRA) critique, part 1: metabolism (p. 721). - Extrapolating from occupational cohorts to environmental exposures is “likely to underestimate risk by 3‑ to 9‑fold” because metabolism below 1 ppm is “more efficient, and alternative” (Kim et al., 2006a, b). - Metabolite formation is nonlinear, with saturation “around 1 ppm”. Workers below 0.1 ppm metabolise benzene to active metabolites “about nine times more efficiently”. - People exposed to ambient air well below 1 ppm also use a “previously unrecognized” pathway (Rappaport et al., 2009).

QRA critique, part 2: the Pliofilm cohort (p. 721). - The NIOSH Pliofilm study of the 1970s has been “widely used by government agencies and others” for benzene–leukaemia dose–response: OSHA 1987, EPA 1998, IARC 1982, Paxton, Silver, Vlaanderen and others. - “More recent analysis” suggests the published data miss “approximately 4–9 additional leukemia deaths” over 1950–1996 follow-up (Infante 2013a, b). Both papers were only “submitted for publication” (p. 722).

Conclusion (p. 721). Taken together, the metabolism and missing-deaths findings indicate that “quantitative cancer risk assessments have underestimated cancer risk resulting from environmental benzene exposure levels”.

Appraisal. - The argument has four strands: (a) a wider range of cancers than regulatory consensus accepts; (b) effects at lower exposures; (c) vulnerable windows (before birth, childhood); (d) systematic bias in QRA from nonlinear metabolism and from a flawed canonical dataset. - (b) and (d, metabolism) rest on peer-reviewed work by a major research group (Rappaport, Rothman, Smith, Lan and colleagues). (d, missing deaths) rests on the author’s unpublished work. - Contestation visible only in the reference list. One reference is Lan et al. (2006), “Response to Lamm and Grunwald” in Science (p. 722), so the low-dose blood-toxicity finding was publicly challenged. The text does not mention the challenge. Crump (1996) and Paxton (1996), reanalyses of the Pliofilm cohort, are also listed (p. 722) but not discussed. - Infante (2011) is titled “The IARC October 2009 evaluation of benzene carcinogenicity was incomplete and needs to be reconsidered” (p. 722). So his wide-spectrum claim went beyond the IARC 2009 evaluation, a gap he had himself argued in print should be closed. The annex text does not tell the reader that IARC’s view was narrower. - The childhood-leukaemia studies rest on exposure proxies (parental occupation, home painting during pregnancy, living near petrol stations), not measured benzene. Infante’s statement that exposures were “likely” well below 1 ppm is an inference, and the proxies also involve other solvents and hydrocarbons. - [External — verify: - IARC Vol. 100F (2009/2012) found sufficient human evidence for AML/ANLL and “positive associations” for ALL, CLL, MM and NHL. IARC Vol. 120 (2017/2018) is reported to have kept “sufficient” for AML in adults and put several of the other cancers under “limited” evidence. - Industry-affiliated reanalyses (e.g. Price et al., 2012) disputed the low-dose “more efficient metabolism” finding as a possible artefact of background levels. - The EU occupational limit for benzene was cut from 1 ppm towards 0.2 ppm, with transition periods (Directive (EU) 2022/431), after ECHA’s risk assessment committee recommended 0.05 ppm in 2018. - IARC evaluated automotive gasoline and some oxygenates in 2025 (Monographs vol. 138). - Check whether the Pliofilm “missing deaths” papers were ever published.]

3. Asbestos – the campaign for a global ban — Castleman and Gee (pp. 724–726)#

Framing (p. 724). Several harmful agents from both volumes “have been banned or phased out at a global level (TBT, CFCs, leaded petrol)”, some with essential-use exceptions (DDT). A global mercury treaty “will be signed in Minamata … in October 2013”. “But what of asbestos, banned throughout Europe in 2005?”

Burden (p. 724). - The WHO estimates at least 107,000 deaths a year worldwide from occupational exposure. “Thousands more” deaths a year are attributed to non-occupational exposure. - The WHO estimates 125 million people are exposed at work, much of it from installed materials in buildings and vehicle brakes. - “The eventual asbestos death toll may reach 5–10 million”. This figure is unsourced in the annex. The WHO figures are attributed to WHO but no specific WHO document is referenced for them (the one WHO reference, WHO 2012, is cited only for the EU bans). - The consumption figures below are also uncited.

Global consumption (p. 724). - Consumption halved from historic levels in the 1990s, then “stabilised at around 2 million metric tonnes per year since 2000”. Russia is the leading producer. - In 2004, Russia, China, India, Kazakhstan, Thailand and Ukraine accounted for about three-quarters of world consumption. Iran, Brazil, Vietnam and Indonesia are other major consumers. - “Bans in over 50 countries have been offset by dramatically increasing use in China, India and other emerging economies.”

Bans and international support (p. 724). - National bans since 2000: all EU countries (WHO, 2012), Chile, Argentina, Australia and Japan. - In 2006 the ILO and WHO called on all countries to ban asbestos and set up a programme to help with national elimination plans. - Thailand: with ILO/WHO help it “move[d] toward banning”. In the process “it was established” that a town-house roof would cost “only USD 65 more” asbestos-free, adding 0.5% to the overall cost (apparently of the house, not of the roof). No source is given for this figure. “However, strong pressure from Russia has held up this progress.” - WHO 2012 (Annex 4) lists substitutes.

Legacy stock (p. 724). - A “vast tonnage” of asbestos-containing material (ACM) remains in buildings, vehicles and industry even where new use is banned. - Some countries require ACM to be inventoried. Removal must be done by trained, certified workers following special procedures. - On 13 March 2013 the European Parliament backed, by a large majority, a report calling for removal of asbestos from all public buildings and buildings with public access by 2028, an EU-wide model for screening and registration, and a removal roadmap.

UK schools (pp. 724–725). - In 2013 Parliament held hearings on asbestos in schools following 139 mesothelioma deaths among school teachers in the last ten years. - 253 teachers have died since 1980, and the rate rose from 3 a year in the 1980s to 14 a year in the last decade (Lees, 2013). - Children are also at risk. School materials are mainly chrysotile, with “large quantities of amosite (brown) and some crocidolite (blue)” (MRC, 1997). - Professor Peto’s evidence: roughly “100 to 150 mesothelioma deaths a year in women now could be from asbestos exposure in schools in the 1960s and 1970s”. If fibre levels are ten times lower now, deaths may fall “by a factor of ten in fifty years”. Under current conditions “there might be 20 or 30 deaths a year in men caused by asbestos exposure in schools” (pp. 724–725). The ellipsis in the printed quotation obscures the context of the men’s figure. - There have been successful compensation claims by people exposed as children at school (UK Supreme Court, 2011) (p. 725). The reference on p. 726 is muddled: it names a Supreme Court judgment and “High Court QBD Liverpool District”. [External — verify: Sienkiewicz v Greif; Willmore v Knowsley MBC [2011] UKSC 10.]

Netherlands limits (p. 725). - Asbestos was banned “20 years ago”, but asbestos still in place “will give rise to exposures for many years to come”. - The Health Council (GR, 2010) recommended lifetime environmental maximum permissible risk limits, measured by transmission electron microscopy (TEM): 2,800 fibres/m³ for chrysotile, 1,300 for chrysotile with up to 20% amphibole, and 300 for 100% amphiboles. - Recommended occupational limits were 2,000, 1,300 and 420 fibres/m³ for 40 h/week over 40 years, corresponding to “4 per 100 000” occupational cancer deaths. - All are “well below” the current Dutch occupational limit of 20,000 fibres/m³ by TEM (0.01 f/cm³, or 10,000 fibres/m³ by phase-contrast microscopy).

Measurement technology (p. 725). TEM gives much higher magnification than phase-contrast optical microscopy (PCM). However, “mainly because of cost and availability considerations”, PCM has been routinely used in workplace sampling, and therefore in the epidemiological dose–response studies and in regulations. TEM “detects even the smallest asbestos fibres” and is considered best for environmental sampling. This is an important mechanism: the cheaper instrument shaped the evidence base and the metric in which regulatory limits are expressed. The annex’s only quantitative handle is the Dutch limit’s conversion (20,000 fibres/m³ by TEM = 10,000 by PCM), i.e. TEM counts about twice as many fibres at that level. The annex does not claim that the gap between the recommended and current Dutch limits is caused by the instrument; the recommended values come from the Health Council’s risk-based derivation.

World Bank Good Practice Note (WBG, 2009) (p. 725). - Asbestos-containing material “should be avoided in new buildings”. Existing facilities should have an asbestos management plan: locations, condition, monitoring, access procedures and staff training. - On substitutes: fibre-cement roof panels cost 10–15% more to manufacture; the newer polypropylene-cellulose-cement roofing costs about 12% more and “has superior impact resistance”; non-asbestos fibre-cement panels are lighter, less brittle and have better nailability. The premium is “to some degree offset” by avoided hygiene, hazard and disposal costs. Micro concrete tiles are “cheaper than A‑C to produce”, and iron pipes outperform asbestos-cement pipes.

Corporate policies (p. 725). Honeywell, Dow, ICI and Unilever have “global policies of not using asbestos in new construction” and follow codes of practice for their own buildings. [External — verify: Honeywell and Dow (via Union Carbide) carried substantial asbestos liabilities, which is plausibly part of the reason. The annex does not say this.]

Canada (pp. 725–726). - Canada was the world’s largest exporter when it challenged the French/EU ban at the WTO in 1999; France and the EU won (see Vol. 1). [External — verify: WTO records date Canada’s request for consultations (DS135) to May 1998, with Panel and Appellate Body reports in 2000–2001.] - Continued Canadian mining, almost all for export to India and other Asian countries, “has drawn strong opposition” from Canadian and other public health bodies. A CBC documentary (2009) showed Indian workers carrying Canadian asbestos “loosely in their arms, the dust swirling around their faces” (p. 725). - Rotterdam Convention. Chrysotile listing would require exporters to notify importers in advance and obtain prior informed consent. The Chemical Review Committee recommended it, but Canada “led opposition from 2004”. In 2011 Canada “blocked the addition just after India announced it would drop its opposition”. Later in 2011 the last operating Canadian asbestos mine closed (p. 726). - Jeffrey Mine. In mid-2010 Quebec offered a CAD 58 million loan guarantee to foreign investors, mainly Indian, to reopen the Jeffrey Mine in the town of Asbestos. - Opponents included unions, opposition parties, the Canadian Cancer Society, and leading medical, public health and human rights associations. A delegation of Asian unionists, victims and activists came to Quebec. - Media reports showed that asbestos trade associations outside Canada “closely followed the lines developed by industry defenders in Canada”. - The Indo-Canadian businessman promoting the reopening was “rebuked publicly by the World Health Organization for misrepresenting WHO policy that no exposure for asbestos was free from cancer risk”. - Deadlines slipped repeatedly amid adverse French- and English-language media coverage (p. 726). - Federal politics. Opposition parties advanced legislation to ban asbestos and assist displaced workers. On 4 September 2012 Quebec’s newly elected government cancelled the public loan (as printed; [External — verify: 4 September 2012 was the election date, and the cancellation may have been announced later that month]). Ottawa “indicated” (undated in the annex) that it “will no longer fight” listing chrysotile under Rotterdam (p. 726).

Networks and courts (p. 726). - Conferences advanced cooperation, e.g. Rome 2006. Osasco, Brazil (2000) drew 32 countries and was “followed by the first national asbestos bans in South America”. - The Asian Asbestos Network brings together health professionals, unionists and victims’ groups. - Brazil: a federal ban failed, but populous states enacted bans, which the Supreme Court upheld in 2008. Labour inspector Fernanda Giannasi blocked export shipments through the port of Santos (São Paulo), leading to court battles. - Since 2000, unions, community groups and international health-professional bodies (e.g. JPC-SE, 2012) have stepped up efforts. But “the road ahead remains difficult”, especially in India, China and Russia.

Appraisal. - This is primarily a status report written by campaigners. It is rich on political-economy mechanisms: the hazard moving to less-regulated markets through trade; producer states blocking an information-sharing treaty; state-backed finance for mines; industry messaging copied across borders; misstating an international health body’s position; international law (WTO, Rotterdam) as an arena; civil society networks; media exposés; elections; sub-national action; individual officials; courts. - The cost-of-substitutes evidence comes from the World Bank note (quoted at length) and the Thailand figure (unsourced, arising from the ILO/WHO-assisted process). Both come from bodies that had already endorsed moving away from asbestos, so they are credible but not neutral. - The mortality projection is unsourced. Peto’s figures are expert opinion given to Parliament. - [External — verify: - Canada banned asbestos from the end of 2018. - Chrysotile has still not been listed under Rotterdam Annex III, with Russia and other producers blocking. - The US EPA banned ongoing chrysotile uses in 2024. - Brazil’s Supreme Federal Court struck down the federal chrysotile law in 2017. - Directive (EU) 2023/2668 cut the EU occupational limit and moves to electron microscopy. - The Minamata Convention was signed in October 2013 and entered into force in 2017. - Later global burden estimates are well above 107,000 deaths a year.]

4. PCBs and the precautionary principle — Koppe (p. 727)#

Opening (p. 727). Vol. 1 ended with the statement that “the non‑application of the precautionary principle has left us with a legacy, the total effects and costs of which can only be guessed at”. In 2000–2010, “parts of the effects on human health and the costs became clearer”.

Diabetes (p. 727). - “Important papers” link PCBs with diabetes, especially type 2 (Lee et al., 2006). The only paper cited is titled as a dose–response relation between serum persistent organic pollutants and diabetes (p. 727), i.e. POPs in general, not PCBs specifically. - There is a “sharp rise” in the disease: 1 million of the Netherlands’ 17 million people have it. The sentence is garbled in the source: “and the same numbers are mentioned”. - Costs in the Netherlands are “about EUR 4.5 billion every year”. - No fraction of diabetes or its cost is attributed to PCBs. The juxtaposition implies a PCB-caused share but does not show one.

Epigenetics (p. 727). - Definition: changes in gene expression without DNA sequence change, via DNA hypo- or hypermethylation or histone changes. - Prenatal and early-life factors “like PCBs … appear to play a role” in chronic diseases such as diabetes. - “PCBs are causing intra‑uterine growth retardation.” Follow-up shows more obesity (Valvi et al., 2011). The only citation follows the obesity sentence, and its title concerns prenatal PCB/DDE/DDT and overweight in children, so the growth-retardation claim is effectively uncited. - Pdx1 (a transcription factor essential for pancreatic beta cells) is lower in growth-retarded offspring in a rat model with reduced beta-cell mass. In adult rats that develop diabetes it is “permanently silenced” (no citation; the annex does not say that the model involved PCB exposure). - The Dutch Hunger Winter of 1944–45 produced more growth-retarded babies, with more adolescent obesity and adult diabetes and epigenetic changes (no citation). This is an analogy from famine, not from chemical exposure.

Slovakia, Michalovce “hot spot” (p. 727). - Follow-up of children found abnormal gene expression, including genes for insulin regulation, mostly reduced in relation to PCB exposure. - Two groups of children, one with high PCB and one with higher DDT/DDE/HCB exposure, showed “the same 14 genes were epigenetically changed in the same direction”. Koppe reads this as indicating “the additive effect” of two chemical classes. - One of the genes, TSGA14, is “related to familial autism” and to obesity in mouse studies (Mitra et al., 2012). - Notes: (a) the Mitra reference title describes a “toxicogenomic” gene-expression analysis in “Slovakian girls” (p. 727), so calling the genes “epigenetically changed” may go beyond what was measured, and the text’s “children” does not say the sample was girls only. (b) Two separately exposed groups showing the same changes indicates a shared target; it does not by itself show additivity, which would need data on combined exposure. The additivity reading is Koppe’s inference.

Conclusion (p. 727). “Denying toxic effects of PCBs has not only made more people ill, but has also hampered the innovative development to cure the epigenetic changes.” And: “This denial is still taken place inhibiting new innovations and developments.”

Appraisal. - This is the thinnest update: three references, with key claims uncited. - The inference chain crosses different models: a growth-retardation rat model, a famine cohort, and a gene-expression study in Slovak children. - Lee et al. (2006) is cross-sectional. [External — verify: in that US population study the strongest associations were with some organochlorine pesticides rather than PCBs, and the design leaves reverse causation open.] - The “denial” claim names no actor, mechanism or evidence, and the innovation-harm claim is asserted. - The most analytically useful element is the possible action of chemically distinct agents on the same genes, which, if effects combine, would challenge chemical-by-chemical assessment. Even that rests on one study, described loosely, and the additivity step is inferred (see above). - [External — verify: IARC classified PCBs as Group 1 (carcinogenic to humans) in 2013 (Monograph 107). The Stockholm Convention targets to eliminate PCBs in equipment by 2025 and manage PCB wastes by 2028 are widely reported as off track.]

5. Ozone layer update — summary of WMO 2006 “Twenty Questions” (p. 728)#

Science summary (p. 728). Standard atmospheric chemistry: catalytic chlorine/bromine cycles, in which one atom destroys “many hundreds” of ozone molecules; polar stratospheric clouds driving the Antarctic hole. Key numbers: severe Antarctic depletion first seen in the early 1980s; seasonal (August–November), peaking in early October, with overhead ozone cut “by as much as two‑thirds” at some locations; in most years the hole’s maximum area “far exceeds the size of the Antarctic continent”; global average about “4 % lower” than in 1980, beyond natural variability, with loss very small near the equator and increasing toward the poles.

Policy effect and outlook (p. 728). - “As a result of the Montreal Protocol”, the total abundance of ozone-depleting gases “has begun to decrease”. If provisions are followed, the decrease continues through the century. - Some gases (halons, HCFCs) “are still increasing” but will decline “in the next decades if compliance … continues”. - Around mid-century, effective abundance “should fall to values that were present before the Antarctic ‘ozone hole’ began to form in the early 1980s”. - Surface UV has increased where depletion is observed. - Ozone depletion “is not the principal cause of climate change”, but the two are linked. Both ozone-depleting gases “and substitute gases contribute to climate change”. - “Substantial recovery” is expected near mid-century “assuming global compliance”. Climate change could speed or delay recovery, and volcanic eruptions could temporarily reduce ozone.

Appraisal. - This is the only update whose central story is a response that is working (other pieces report partial gains, e.g. national asbestos bans and early growth-promoter bans), and it is purely descriptive. It says nothing on precaution, on Farman’s own discovery, on industry resistance, on the lag between the 1970s warnings and the 1987 Protocol (external context, not in the annex), or on why this case succeeded. - The editors do not use it as a counterweight to their “harm expands” thesis. Yet it carries two lessons that the thesis does not capture: harm continues for decades after action (the decline of ozone-depleting gases continues “throughout the 21st century”), and recovery is conditional on continued compliance. - The source was already dated. [External — verify: a 2010 update of Twenty Questions was published around 2011.] - It notes that substitute gases also contribute to climate change. That is a substitution trade-off, though the summary states it as a fact and does not frame it as a problem. - [External — verify hindsight: the Kigali Amendment (2016) phases down HFCs. Unexpected CFC-11 emissions detected in 2018 were traced largely to eastern China and then fell. The 2022 WMO/UNEP assessment projected return to 1980 values around 2066 over Antarctica, 2045 in the Arctic and 2040 elsewhere.]

6. DES: the view from 2013 — Swan (pp. 729–730)#

Framing (p. 729). Looking back at the 2000 chapter, “the long‑term consequences we wrote about in 2000 were merely the tip of the iceberg”. - DES is “now viewed as the archetypical environmental estrogen and endocrine disruptor”. Understanding of how it disrupts fetal reproductive-tract development has grown: it leads “to cancer in rare instances but reproductive dysfunction commonly”. - Rodents and humans show “a remarkable concordance” in outcomes (Newbold, 1995). Follow-up shows impacts “far more extensive and far‑reaching than had been anticipated”. - “Perhaps most alarming”: rodent data show that exposure in the womb “causes heritable epigenetic alterations” (no specific citation). Studies of DES “grandchildren” are “starting to identify third generation effects in humans”.

Daughters (p. 729). - Most data come from three NCI cohorts in the US with known prenatal exposure and unexposed controls. - Hoover et al. (2011) compared 4,653 women exposed in the womb with “1 027” unexposed women. [External — verify: I recall the NEJM paper reporting about 1,927 unexposed controls. The annex figure may be a transcription error.] - It found increased risks of clear-cell adenocarcinoma of the vagina and cervix “and 11 other more common adverse health outcomes”: infertility, a range of adverse pregnancy outcomes, early menopause, and breast cancer in women over 40. Hazard ratios ranged “from 1.4 to 8.1”. - For most outcomes, women with vaginal epithelial changes (VEC) at a young age, “a marker of higher DES dose and exposure early in gestation”, had far greater risk than exposed women without VEC. In the VEC group, cumulative risks of infertility and preterm birth were 68% and 41%, versus 15% for both in the unexposed. - “Becoming pregnant at all, maintaining the pregnancy, and delivering a live term baby has proven to be a severe challenge.”

Sons (p. 729). - Palmer et al. (2009), three US cohorts: significant excess cryptorchidism, epididymal cyst and testicular inflammation or infection, with stronger associations when exposure began before the 11th week. - A meta-analysis of three testicular dysgenesis endpoints (cryptorchidism, hypospadias, testicular cancer) found “doubled risk ratios” after DES (Martin et al., 2008).

Grandchildren (p. 729). - The DES grandchildren are “survivors of high‑risk pregnancies”, followed most extensively in the NCI third-generation cohort. - Based on the mothers’ reports, overall birth defects were elevated in sons (OR 1.53; 95% CI 1.04–2.23) and daughters (OR 2.35; 95% CI 1.44–3.82). However, “the pattern of defects did not resemble” that seen in the exposed generation (Titus-Ernstoff et al., 2010). - Few reproductive-tract anomalies were reported, “perhaps because most participants had not reached reproductive age”. “Further data is needed.”

Conclusion and BPA warning (p. 730). - “The tragic DES ‘experiment’ had consequences even more profound and far‑reaching than we predicted in 2000.” This is “an important lesson” for synthetic oestrogens now on the market. - Sir Charles Dodds, who synthesised DES, “recognized the estrogenicity of bisphenol‑A (BPA) in 1930”. Its use in plastic began in 1940 and “soared”. Today “10 billion pounds” are made each year. - Exposure of pregnant women “is ubiquitous and studies to identify its effects are only now being conducted”.

Appraisal. - This is the strongest empirical update in the annex. It rests on large NCI cohorts with documented exposure and controls. Dose and timing gradients (VEC; exposure before week 11) strengthen causal inference, and rodent–human concordance adds coherence. - Swan is appropriately cautious on the third generation: the defects are mother-reported, the pattern is different, and the data are early. - The annex reports twelve outcomes with raised risk but does not say how many outcomes were examined, so the scope for chance findings across multiple comparisons cannot be judged from the page. The dose/timing gradient “for most outcomes” argues against chance. - The BPA dates look loose. [External — verify: Dodds and Lawson’s paper on the oestrogenic activity of such compounds dates from 1936, and polycarbonate and epoxy uses developed in the late 1940s–1950s.] The substantive point stands: decades passed between knowing BPA was hormonally active and systematically studying the effects of exposure. - [External — verify hindsight: EFSA (2023) cut the BPA tolerable daily intake by about 20,000-fold, and the EU banned BPA in food-contact materials (Reg. (EU) 2024/3190). Third-generation DES studies after 2013 reported further outcomes (e.g. ADHD, hypospadias).]

7. Antimicrobials as growth promoters — Silbergeld (pp. 731–734)#

Introduction (p. 731). - Silbergeld endorses the Vol. 1 authors’ description of the practice as “resistance to common sense”. “Much has been little changed (to quote Edmund Burke)” in policy or drug use. - Countries that banned non-therapeutic uses early “have generated convincing empirical evidence” that the bans did not reduce productivity, did not measurably affect sector economics, and did not harm animal health or welfare. - In Denmark and the Netherlands, bans were “accompanied by improvements in animal husbandry”, partly to prevent impacts and partly in response to “separate public pressures”. - These gains were “overwhelmed” by major producing countries failing to act and by rapid global uptake of “the industrial model of food animal production, along with GPA use”. - The US “has restricted use of only one drug (fluoroquinolones in poultry feed)” and “retains its resistance to change”. - “The US government finally compelled the release of industry data, which showed that nearly 80 % of antimicrobial production is used as feed additives for food animals.” - “Overall drug uses have increased substantially as Brazil, China, and India are now among the top 15 producers of poultry, and China is one of the top producers of swine” (OECD-FAO, 2012). The citation supports the production rankings; the increase in drug use is inferred from them, not measured. The “challenge now is to communicate the evidence” to this “new wave”. - Thomas O’Brien is quoted twice on the same page, an editing duplication: the use of antimicrobials anywhere “can increase resistance to any antimicrobial anywhere else” (O’Brien, 2002).

The resistance crisis (p. 731). - Scientists discuss “the end of the antibiotic era” (Bush et al., 2011). - The “single most important cause of infectious morbidity and mortality in the US in 2012” is no longer HIV or TB but MRSA. This claim is uncited. - In the EU, livestock (especially swine and cows) are “a major source of human exposures to MRSA” (also uncited at this point). - Hence: “this is no longer an issue of a failure to heed ‘early warnings’ but rather a case study of intransigence in the face of overwhelming evidence”. Success “requires us to understand the sources and motivations of the continued opposition”.

New evidence on links to human health (pp. 731–732). - It was “quickly admitted” that feed use selects for resistance in animal gut bacteria. But “many industry and government scientists have argued for decades” that there is no evidence of increased human risk. That position is now “increasingly difficult” to hold against “overwhelming scientific evidence” (Marshall and Levy, 2011). - Monitoring after removal from feed in northern Europe showed “significant reductions in avoparcin resistance in enterococci” from animals, food products and human samples over the first three years. - Since 2002, molecular methods (multilocus sequencing, whole-genome sequencing) test links between swine farms and MRSA, and between poultry houses and C. difficile. For a public “used to TV shows like CSI”, this “is readily understandable” (p. 732). - “Studies have been conducted in many countries on the prevalence of drug resistant pathogens in the food supply and in farms using the standard industrial model” (no findings or citations given). Aquaculture uses antimicrobials heavily worldwide, and there are exposure routes besides food (uncited here; Feingold et al., 2012 is cited for non-food pathways a paragraph later).

“One reservoir” (p. 732). - She criticises “the old notion” of judging resistance-selecting actions by the clinical importance of the drug and the severity of disease caused by the bug (Silbergeld et al., 2008). - Stuart Levy and others “had warned against the dangers of this concept in the early 1970s” because bacteria exchange resistance genes horizontally across species and strains. These fundamentals “were apparently ignored by regulators and industry in their protection of drug use in animal feeds”. - Awareness has grown of environmental impacts that expand the “resources of resistance” available to microbes (Davis et al., 2011), and of non-food exposure routes (Feingold et al., 2012).

Economics (pp. 732–733). - “Industry has long argued — since 1947 — and government has long accepted” that growth promoters confer productivity advantages. On that basis, Vol. 1 treated a ban as a risk/benefit balance. - Evidence now challenges the benefits. Review of the early studies behind FDA registration shows their limitations (no citation). “The most persuasive evidence actually comes from two recent industry studies as well as the careful documentation of empirical experience of the industry in Denmark” (p. 732). - Inconsistency: only one of the two studies then described is an industry study (Perdue). The other is a USDA analysis of a government survey. The Danish documentation is mentioned but neither cited nor described. - Perdue study (1989; published Engster et al., 2002). Perdue is described as “one of the oldest and largest poultry integrators in the US”; the study covered its growers in Maryland and North Carolina. - It had “many of the hallmarks of a randomized clinical trial”: size, duration, treatment and control groups, crossover design, direct measurement of growth, feed consumption, disease, mortality and processing acceptability. The only variable was the standard drug package. - Results showed no “major impacts” on any economically important parameter (p. 732). - No variance data were reported, so one cannot conclude there was no effect. But effects were “relatively small and well below” earlier claims of “double digit increases in growth rates”. - There was no support for reduced disease or mortality (p. 733). - Graham et al. (2007), a reanalysis in which Silbergeld was a co-author (“We reanalyzed”), monetised marginal benefits and costs. Importantly, it asked the question “assuming a significant increase in benefit (improved performance)”, so the result does not depend on the drugs having zero performance effect: the value of the gain was outweighed by the cost. They found that “instead of an economic benefit, the use of AMs in animal feeds was actually a net loss for producers”. “There has been no dispute of this paper in the published literature” (p. 733). - USDA swine survey (2004 data; McBride et al., 2008): - The design is weaker: self-reported, non-random and observational. - A “‘surprising result’” in the authors’ words: antimicrobial use was “significantly and negatively associated with productivity”, whereas carefully managed nutrition was positively associated (p. 733).

“How could this be so?” (p. 733). - These failures to confirm the benefits “should cause us to demand a compete [sic] re‑evaluation”. - The debate “has pitted much of the agribusiness sector and many national agricultural ministries against public health practitioners and public health agencies”. - Early studies were inadequate. As late as 1987 industry published a study of groups of “7 new born chicks” tested “in a lab for 9 days” (Feighner and Dashkevicz, 1987), from which it is “difficult to infer anything” about real flocks. - Historical rationale. Silbergeld concedes: “There may be reason to accept the possibility that GPAs once conferred real benefits to industrial food animal production.” Antimicrobials were first shown to substitute for the “animal protein factor”, later identified as vitamin B12. The search for alternatives was driven by the cost of isolating B12 in the 1940s (Stokstad and Jukes, 1959). B12 is now synthesised and present in most feeds, “which may remove the need for additions of drugs”. Advances in breeding, husbandry and feed formulation “may have replaced the need for drugs”.

Conclusions (p. 733). - Use “has very little support in evidence for efficacy and an increasing body of evidence of harm to public health”. - “Since the shoe is clearly now on the other foot, governments and industry have little justification to resist change.” - “This is no longer an issue of ‘early warnings in the absence of proof’, but rather persistence in a clearly dangerous practice in the absence of benefit.”

Appraisal. - Strengths. The benefit side is scrutinised with the same rigour usually reserved for risks. The main benefit study was industry-originated and well designed. The “one reservoir” argument rests on well-established microbiology. The post-ban monitoring data are a genuine natural experiment. - Weaknesses and overstatements: - The central economic conclusion comes from the author’s own reanalysis, and “no dispute in the literature” is not confirmation. - The MRSA claim is uncited and, as phrased (“single most important cause of infectious morbidity and mortality”), very broad. [External — verify: the widely cited comparison at the time (Klevens et al., 2007) was that estimated US deaths from invasive MRSA exceeded deaths from AIDS; that is narrower than the annex’s wording.] - The “80 %” figure is framed as “feed additives”. [External — verify: the FDA sales data it likely rests on covered antimicrobials sold for food animals for all purposes (including therapeutic use and ionophores not used in human medicine). Comparing them with human-use sales by weight has been criticised.] - The US “only one drug … in poultry feed” statement looks imprecise. [External — verify: poultry fluoroquinolones were withdrawn in 2005 and were administered in water. FDA also restricted extra-label cephalosporin use in 2012 and issued Guidance #209 in 2012.] - [External — verify] The EU-wide ban on antibiotic growth promoters from 2006 is not mentioned. That omission makes “much has been little changed” look more pessimistic for Europe than the record supports. - The claim of no harm to animal welfare in Denmark is stronger than some assessments. [External — verify: the WHO’s 2002 review of Denmark’s termination reported increased therapeutic use and some problems in weaner pigs, alongside large falls in total use and little effect on productivity.] - On McBride et al. (2008): [External — verify: later USDA ERS work (e.g. Sneeringer et al., 2015) estimated small positive productivity effects of roughly 1–3%. That would still support “small benefit” rather than “net harm”.] - [External — verify hindsight: FDA Guidance #213 (final 2013) ended production uses of medically important antimicrobials in the US by January 2017. EU Regulation (EU) 2019/6 restricts prophylaxis and extends import conditions. Global veterinary antimicrobial use is projected to grow in emerging producer countries.]

8. A brief comment on MTBE — Belpoggi (p. 735)#

Claims (p. 735). - “Experimental evidence demonstrates that methyl tert‑butyl ether (MTBE) is carcinogenic in rats and mice.” MTBE “is probably carcinogenic for humans”, “children are particularly at risk”, and “syncancerogenic effects must be taken in consideration”. - In the CMCRC/RI model, MTBE induces “lymphomas/leukemias in females and testicular tumors”. - Human incidence of lymphomas and leukaemias is increasing, “recently … confined to women” (American Cancer Society, 2006; NCI, 2005), and testicular cancer is increasing in many countries (NCI, 2007).

EEA and action (p. 735). “The European Environment Agency calls the case of MTBE one of its Late lessons”, recommending “additional research and extreme precaution” for “this persistent chemical” (EEA, 2001). “While we await new experimental data to better quantify the carcinogenic risks of MTBE, we must recognize the existence of this risk and act accordingly to protect public health.”

The lab’s track record (p. 735). - “In the past, our laboratory had different experiences in which its experimental results were ignored for risk assessment, but then confirmed by epidemiology.” - Example: in 1989 the lab published the first results showing formaldehyde to be a strong multi-site carcinogen, particularly for leukaemias. In 2009 an IARC Working Group found “sufficient evidence in humans for a causal association of formaldehyde with leukaemia” and classified formaldehyde as Group 1 (IARC, 2012). - “Every molecule of MTBE is transformed in one of formaldehyde.”

Appraisal. - No study is cited for the central claim that MTBE causes cancer in rodents, not even the lab’s own publications. The five references are cancer statistics, the EEA’s 2001 report and IARC 100F. - The supporting arguments are indirect: - ecological trends in human cancer rates (she pairs the lymphomas/leukaemias seen in female rodents with a human rise “recently … confined to women”), which could have many causes; - a metabolite argument (MTBE yields formaldehyde), which says nothing about dose [External — verify: or about the body’s own endogenous formaldehyde]; - the lab’s past vindication on a different chemical. - Reframing. [External — verify: the Vol. 1 chapter’s main concern was MTBE’s persistence and mobility in groundwater (taste and odour contamination) and its adoption as a lead substitute without adequate assessment.] Belpoggi moves the case’s centre of gravity to carcinogenicity. The EEA’s own Vol. 1 wording (“extreme precaution”) cannot be checked from this annex. - [External — verify: - IARC classified MTBE as Group 3 (not classifiable) in 1999. - Aspects of the Ramazzini bioassay pathology were later disputed (see Authors and standpoint). - US MTBE use largely ended in the mid-2000s, driven by groundwater contamination and liability rather than carcinogenicity findings. - IARC re-evaluated MTBE with gasoline oxygenates in 2025 (vol. 138). My recollection is an upgrade to Group 2B; check. - The formaldehyde–leukaemia causal link remains debated on grounds of biological plausibility.] - The formaldehyde precedent is a legitimate illustration of a real lag: 1989 laboratory results, 2009 IARC human-evidence conclusion. But the narrative that the lab was “ignored … then confirmed” is the lab’s own account.

9. Hormones as growth promoters — Christoforou (pp. 736–737)#

Status (p. 736). - The WTO Appellate Body report in United States, Canada — continued suspension of obligations in the EC — Hormones Dispute was published on 16 October 2008. It was the AB’s “longest report to date”, and it “reverses most of the conclusions previously reached by the WTO Panel”. - Footnote 1: “formally two reports i.e. AB-2006-5 and AB-2006-6. References are to the paragraphs in Appellate Body report AB-2008-6.” The first sentence is inconsistent with the “AB-2008-5 and AB-2008-6” in the heading and with the footnote’s own second sentence; the “2006” looks like a typo. All paragraph numbers below refer to AB-2008-6.

Headline findings as characterised (p. 736). The AB “confirms practically all the SPS … related arguments made by the Community”. It found the Panel committed “numerous flaws”: - breaching the EC’s due process rights; - failing to make an objective assessment; - “incorrectly reversing the burden of proof”; - failing to apply the correct standard of review; - wrongly finding the import ban inconsistent with the SPS Agreement.

Procedural findings (p. 736). - The report “clarifies the existing dispute settlement rules and provides a clear roadmap for resolving implementation disputes”. - The AB agreed with the EC on which procedure applies to implementation disagreements. - It disagreed on when retaliatory sanctions must cease, and “consequently considers that the United States and Canada did not act in a unilateral manner”. This is a loss for the EC, candidly noted.

Background (p. 736). - After five years’ work, Directive 2003/74/EC (14 October 2003) implemented the Dispute Settlement Body recommendations of 13 February 1998. - After new risk assessments it imposed a definitive prohibition on oestradiol-17β. Evidence on testosterone, progesterone, trenbolone acetate, zeranol and melengestrol acetate was “insufficient” for a definitive conclusion, so a provisional prohibition applied to those five. - The EC notified the WTO and asked the US and Canada to withdraw sanctions. “They refused.” The EC therefore “had no option” but to bring proceedings. The Panel reached “contradictory findings”, and all parties appealed.

SPS analysis (pp. 736–737). - Expert independence. The Panel infringed due process because appointing and consulting two scientific experts “was likely to affect or give justifiable doubts as to their independence or impartiality”. This “compromised the adjudicative independence and impartiality of the Panel”, in breach of DSU Art. 11 (paras 481–482) (p. 736). [External — verify: the concern related to the two experts’ prior involvement in the Codex/JECFA evaluations of the same hormones.] - Restrictive notion of risk assessment. The Panel’s interpretation repeated the “restrictive notion of risk assessment” the AB had rejected in the original case (para 542). This concerned “risk of misuse and abuse in the administration of hormones” (paras 543–555) (p. 737). This implies risk assessment should cover real-world conditions of use, including misuse, not only correct use. - Burden of proof. The Panel erred in allocating the burden in assessing Directive 2003/74/EC under Art. 5.1 (paras 576–584), and allocated it “to the detriment of the Community” (paras 716–718) (p. 737). - Role of the adjudicator. The Panel went beyond its role and “somewhat peremptorily decided what it considered to be the best science, rather than following the more limited exercise that its mandate required” (para 612) (p. 737). - Article 5.7 (provisional measures). The AB reversed the finding that the provisional ban on five hormones failed Art. 5.7 because evidence was not “insufficient” (p. 737). - International standards. Their existence “is not dispositive on the sufficiency of the relevant scientific information”. A member choosing a higher level of protection “is not obliged to frame the scope and methods of its risk assessments in the same manner as the international body” (para 685) (p. 737). - “Critical mass”. The Panel’s “critical mass” standard for insufficiency under Art. 5.7 “was too inflexible” (paras 703–712) (p. 737). [External — verify: the Panel had required a critical mass of new evidence that called into question the fundamental precepts of previous knowledge.]

Outcome (p. 737). - The report is “a significant victory to the Community in substance”. But because of “serious flaws in the Panel’s analysis and the highly contested facts”, the AB “could not complete the analysis”. - Confirming the ban’s consistency and removing sanctions “will need either an agreement between the parties … or another round of litigation”. The EC’s position is “significantly strengthened”, and the SPS findings will be “highly relevant” for its other SPS disputes.

Appraisal. - This is a partisan but reasonably candid legal summary. It records the EC’s losses (on sanctions and unilateralism) as well as its wins, and admits that the AB “could not complete the analysis”, so the WTO-consistency of the EC ban was left unconfirmed. - Its lasting significance for the project is the set of legal standards governing how science is handled when a precautionary measure is challenged: expert independence, burden of proof, standard of review (a court reviewing the regulator’s reasoning rather than choosing the “best science”), the relevance of misuse to risk, the non-dispositive status of international standards, and a flexible reading of “insufficient” evidence. - It says nothing about the science of the hormones themselves. - [External — verify hindsight: the dispute was resolved by negotiation, not litigation: a 2009 EU–US Memorandum of Understanding (and a 2011 one with Canada) opened a duty-free EU quota for high-quality hormone-free beef in exchange for phased removal of sanctions. The EU ban remains.]

Case timeline#

Because this is a multi-case annex, the timeline gives only the dates the annex itself supplies. Lags are computed only where the annex gives both ends.

The authors’ own lessons and conclusions#

Lessons derived from evidence (as stated)#

• Editors: harm from most agents “expands over time” to new kinds of harm, often at exposures once considered “safe” (p. 717).
• Ruzzin: background POP levels “can fully abolish” fish’s nutritional benefits in animal models. The health benefits of fish “may be over‑estimated” (p. 718).
• Infante: effects occur at progressively lower exposures and across more cancer types. Quantitative risk assessments have “underestimated” environmental cancer risk because of nonlinear metabolism and missing cases (pp. 720–721).
• Castleman and Gee: bans in over 50 countries have been “offset” by growth elsewhere. Substitutes are available at modest extra cost. Asbestos already in place continues to cause exposure and deaths decades after bans (pp. 724–725).
• Koppe: PCB health effects and costs “became clearer”, notably diabetes via early-life epigenetic mechanisms, with possibly additive effects across chemical classes (p. 727).
• Ozone summary: the Montreal Protocol has begun reducing ozone-depleting gases. Recovery near mid-century is conditional on compliance, and substitute gases contribute to climate change (p. 728).
• Swan: DES consequences were “merely the tip of the iceberg”: more outcomes, both sexes, a dose/timing gradient, possible third-generation effects (pp. 729–730).
• Silbergeld: growth promoters show little efficacy. Early-ban countries suffered no productivity loss. The resistance-risk framework ignored horizontal gene transfer. The debate has moved from early warnings to “intransigence” (pp. 731–733).
• Belpoggi: the lab’s formaldehyde results were confirmed after being ignored, which implies its MTBE results deserve action now (p. 735).
• Christoforou: the AB established that panels must respect expert independence, allocate the burden of proof correctly, not pick “the best science”, and not treat international standards as dispositive or “insufficiency” as requiring a rigid “critical mass” (pp. 736–737).

Recommendations and advocacy (as stated)#

• Fish: “urgent need” to control pollutants in seafood and to study health impacts of current farmed fish (p. 718).
• Asbestos: the title itself calls for a global ban. The piece implicitly endorses ILO/WHO elimination plans, the European Parliament’s 2028 removal target, and chrysotile listing under Rotterdam (pp. 724–726).
• MTBE: “we must recognize the existence of this risk and act accordingly” (p. 735).
• Antimicrobials: “demand a compete re‑evaluation”. Governments and industry “have little justification to resist change”. Communicate evidence to the “new wave” of producer countries. Study the motivations of the opposition (pp. 731, 733).
• DES: treat the DES experience as “an important lesson” for synthetic oestrogens now on the market, especially BPA (p. 730).
• PCBs: implicitly, stop “denial” to enable innovation in curing epigenetic changes (p. 727).
• Hormones: none explicitly. The piece argues that the EC position is legally strengthened (p. 737).
• Ozone and benzene: no explicit recommendations. The ozone summary is descriptive; Infante’s piece implies that quantitative risk assessments should be revised upward but does not say how (pp. 721, 728).

Mechanisms and dynamics#

1. Expanding harm profiles and falling effect thresholds. - Continued study finds new endpoints, lower-dose effects, susceptible windows and later generations. - Benzene: from fatal blood disease at high dose (1897) to AML/MM and blood toxicity below 1 ppm, and childhood leukaemia from exposure before birth (p. 720). - DES: from rare cancer to common reproductive dysfunction in daughters and sons, with possible third-generation effects (p. 729). - PCBs: from earlier concerns to diabetes and obesity (p. 727). - Asbestos: teachers and schoolchildren (p. 724); recommended Dutch limits roughly 10–50 times below the current occupational limit (p. 725). - Part of this is detection-driven: new tools (biomarkers, molecular sequencing, electron microscopy, long cohort follow-up) reveal what was always there. The annex implicitly shows this for antimicrobials (p. 732), asbestos measurement (p. 725) and benzene metabolism (p. 721). It never considers the flip side: more study of a suspect agent will find more associations, some of them spurious.

2. How risk assessment methods produce systematic underestimation. - Extrapolation across nonlinear biology: high-dose occupational data applied to low environmental exposures, when metabolism is more efficient at low dose (p. 721). - Canonical dataset lock-in: one 1970s cohort underpins decades of agency assessments, so errors propagate everywhere (p. 721). - Instrument choice: cheaper optical microscopy undercounts the finest fibres, yet underpins epidemiology and limits (p. 725). - Chemical-by-chemical assessment: it would miss combined action across classes (my inference from Koppe’s additivity claim, p. 727; the annex does not discuss assessment methods here). - Assuming ideal use: the WTO AB faulted a “restrictive notion of risk assessment” that excluded “misuse and abuse” (p. 737). - Wrong unit of analysis: judging resistance risk drug by drug and bug by bug ignores horizontal gene transfer across a shared reservoir (p. 732).

3. Legacy stocks, latency and irreversibility. Harm continues long after action: - asbestos in place in buildings (pp. 724–725), and school exposures in the 1960s–70s producing deaths decades later, with Peto projecting a fifty-year decline (pp. 724–725); - the Netherlands 20 years after its ban (p. 725); - ozone-depleting gases persisting to mid-century (p. 728); - DES effects in grandchildren (p. 729); - persistent organic pollutants in food chains (p. 718).

The time from action to relief runs to decades, and in the ozone case recovery is explicitly “assuming global compliance” (p. 728).

4. Geographic and market displacement. Restrictions in rich jurisdictions push production and use elsewhere: - asbestos bans in 50+ countries “offset” by growth in China and India (p. 724); - Canadian exports almost entirely to India and Asia (p. 725); - the industrial animal-production model, with growth promoters, spreading to Brazil, China and India (p. 731).

Exporting states then block international information-sharing mechanisms (Canada on Rotterdam, 2004–2011, p. 726) and pressure importing states (Russia on Thailand, p. 724). The annex implies a double standard: Canadian output went almost entirely to Asian workplaces, which Canadian public health bodies themselves opposed (p. 725).

5. Political economy: interests, finance and state action. - State finance to sustain a hazardous industry (Quebec’s CAD 58m, p. 726). - Industry messaging propagated across national trade associations (p. 726). - Misstatement of international health bodies’ positions (WHO rebuke, p. 726). - Ministries of agriculture aligned with agribusiness against public health agencies (p. 733). - “Industry and government scientists” maintaining for decades that there was “no evidence” of human harm (pp. 731–732). - Industry data on antimicrobial use released only under government compulsion (p. 731). - A recurring pattern in the annex: governments and ministries that sponsor or promote an industry defend it, and health agencies and professionals oppose it (pp. 726, 733).

6. Unexamined benefit claims. - Growth-promoter benefits were assumed from 1947 on weak evidence (tiny, short lab studies, p. 733) and “accepted” by governments (p. 732). When a well-designed industry study and observational data appeared, benefits were small, absent or negative (pp. 732–733). Silbergeld’s inference is that the burden of justification then shifts: “the shoe is clearly now on the other foot” (p. 733). - Fish health benefits were similarly taken for granted while the product changed underneath the claim (p. 718). - More broadly (my inference; the annex does not discuss incumbents’ cost claims), incumbents’ claims about the product’s benefits and about the cost of alternatives appear to have got less scrutiny than claims of harm. When tested, alternatives proved cheap: USD 65 extra for an asbestos-free Thai town-house roof, 0.5% of the overall cost; a 10–15% manufacturing premium for fibre-cement panels, partly offset downstream; some substitutes technically superior (pp. 724–725).

7. Path dependence and practices that outlive their rationale. Growth promoters began as a cheap substitute for vitamin B12 (1940s). The practice persisted after B12 synthesis and husbandry advances removed the original need (p. 733). This is a clear case of institutional and technical lock-in outlasting its justification.

8. Optimisation targets and unstudied consequences. Aquaculture feed was optimised for fish growth and health, with “important financial resources” spent on it (p. 718). The ratio of the more nutritious omega-3s (EPA, DHA) to n-6 fatty acids in farmed salmon fell about 50%. The change itself was measured (NIFES data), but its health impact “has not been researched” (p. 718). A system optimised on the metrics its owners care about can degrade the properties users rely on; even when the change is recorded, nobody may be tasked with studying what it means for users.

9. Substitution, regrettable and otherwise. - Regrettable: vegetable feeds introduce “new hazardous chemicals” (p. 718); substitute gases for ozone-depleting substances contribute to climate change (p. 728). MTBE was itself introduced as a lead substitute (chapter title cited, p. 735). - Successful: asbestos substitutes are modestly costlier but lighter, less brittle and more impact-resistant, and some (micro-concrete tiles, iron pipes) are cheaper or better (p. 725). - The annex shows both, but draws no general lesson about assessing substitutes before adopting them.

10. How experts and regulators thought (mental models and blind spots). - Resistance risk was framed by the clinical importance of the drug and the severity of disease, despite known horizontal gene transfer, whose implications Silbergeld says were “apparently ignored by regulators and industry” (p. 732). - Regulators and “many people” treated background exposure levels as safe (p. 718). - Adjudicators substituted their own view of “the best science” for review of the regulator’s reasoning (p. 737). - Agencies kept relying on a legacy cohort (p. 721). - Industry’s belief in growth promotion was reinforced by studies designed in ways that could not test real-world performance (p. 733). - The annex does not show the internal reasoning of these actors directly; it infers it from outcomes.

11. Burden and standard of proof. - Explicit in the WTO piece: burden misallocated “to the detriment of the Community” (p. 737), and “insufficiency” not requiring a rigid “critical mass” (p. 737). - Implicit in Silbergeld’s reversal: once benefits are unproven and harms documented, persistence needs justification (p. 733). - Implicit in Belpoggi: act before quantification is complete (p. 735). - The annex shows standards of proof operating asymmetrically. Benefit claims were accepted on weak evidence (p. 733), while harm claims were held to demanding standards: “no evidence” of human risk (p. 732), and the WTO panel’s “critical mass” test (p. 737).

12. Knowledge production and its channels. - Animal data discounted until epidemiology confirms (formaldehyde, p. 735). - Rodent–human concordance (DES, p. 729). - New molecular tools making causal pathways traceable and even publicly intelligible (“CSI”, p. 732). - Monitoring after policy change as a natural experiment (p. 732). - An industry study (Perdue) providing some of the most persuasive evidence against industry claims (p. 732). Silbergeld calls both her key studies “industry studies”, but the second is a USDA analysis. - Author self-citation and unpublished manuscripts as the basis of key claims (benzene, fish, antimicrobials).

13. Law, trade and courts as arenas. - The WTO: Canada vs. France/EU on asbestos, 1999 (p. 725); US/Canada vs. EC on hormones (pp. 736–737). - National supreme courts: Brazil upholding state bans (p. 726); UK compensation for school exposure (p. 725). - Treaty mechanisms: Rotterdam prior informed consent (p. 726); the Montreal Protocol (p. 728); Minamata (p. 724). - Trade law can either constrain or legitimise precautionary measures, depending on how burden of proof, standard of review and “insufficient evidence” are read (p. 737).

14. Civil society, media and politics as change agents. - Transnational coalitions: the Asian Asbestos Network, Osasco (p. 726); health professional societies (JPC-SE, p. 726). - Media exposés: CBC 2009 (p. 725); adverse coverage delaying the mine (p. 726). - Elections: the Quebec 2012 change of government (p. 726). - Individual officials: labour inspector Giannasi (p. 726). - Sub-national action preceding national action (Brazilian states, p. 726). - International agencies: ILO/WHO technical help (p. 724), World Bank lending guidance (p. 725). - Early movers: Denmark and the Netherlands, “separate public pressures” (p. 731).

15. Costs and who bears them. - Workers bear the harm: 107,000 occupational asbestos deaths a year (p. 724); Indian plant workers handling Canadian asbestos (p. 725). - Teachers and children (p. 724). - DES daughters’ fertility and pregnancy outcomes (p. 729), and their children (p. 729). - The general public bears the resistance crisis (p. 731). - Health systems bear chronic disease costs: EUR 4.5 bn a year for diabetes in the Netherlands, though not attributed to PCBs (p. 727). - The presumed benefits went to producers (the annex’s framing of the industry case). Silbergeld argues even that benefit was illusory, a “net loss for producers” (p. 733). If she is right (my inference from here on), the practice imposed public costs without private gain, which makes its persistence a matter of institutional inertia or of benefits to actors other than producers (e.g. drug sellers). The annex does not explore the latter.

16. Innovation effects. - Danish and Dutch bans were accompanied by husbandry improvements (p. 731). - Asbestos substitutes improved on the incumbent in some respects (p. 725). - Koppe asserts that “denial” hampered innovation toward curing epigenetic damage (p. 727). - In aquaculture, innovation (vegetable feeds) solved a resource constraint while creating a nutritional and chemical problem (p. 718). - The annex gives no systematic evidence on innovation, only these illustrations.

17. Framing and language. - Evaluative language runs throughout: “magic” (the Vol. 1 asbestos title, p. 724); “safe” in scare quotes (p. 717); “background levels … many people consider to be at safe levels” (p. 718); “inconvenient truth” (p. 717); “resistance to common sense” (p. 731); “intransigence” (p. 731); “the end of the antibiotic era” (p. 731); “tip of the iceberg” (p. 729); “tragic DES ‘experiment’” (p. 730); “archetypical” (p. 729); “denial” (p. 727); “victory” (pp. 736–737). - Silbergeld explicitly reclassifies her case: no longer “early warnings in the absence of proof” but “intransigence” and “persistence … in the absence of benefit” (pp. 731, 733). The move signals that uncertainty is no longer the operative explanation for inaction. Swan’s “tip of the iceberg” (p. 729) is a related but weaker move: the warning was right, and understated.

Transferable insights (technology-neutral)#

1. Harm profiles tend to widen and effective thresholds tend to fall as study continues. Follow-up with better cohorts, biomarkers and outcome measures finds new endpoints, lower-dose effects, susceptible developmental windows and later generations. - Evidence: pp. 717, 720–721, 729–730, 727, 724–725, 718. - Strength: moderate. Well documented for DES and benzene. Thin for PCBs and fish. The annex’s cases were selected because they turned out harmful, the ozone update is a counterexample, and part of the pattern reflects more research attention.

2. Linear extrapolation from high-exposure data is not automatically conservative. When the underlying biology is nonlinear (e.g. saturable metabolism), low-exposure risk can be underestimated. - Evidence: p. 721 (3–9-fold; about nine times more efficient metabolism below 0.1 ppm). - Strength: moderate. Peer-reviewed mechanistic studies, but the finding was contested (the challenge is visible only in the references, p. 722) and the magnitude is model-dependent.

3. A single canonical dataset can become load-bearing regulatory infrastructure for decades, so its errors or omissions propagate into every downstream estimate. - Evidence: the Pliofilm cohort used by OSHA, EPA, IARC and others; 4–9 possibly missing deaths (p. 721). - Strength: suggestive. Heavy reliance is well documented; the specific error claim rested on unpublished work by an author re-examining his own cohort.

4. The instruments chosen for cost and availability shape what the evidence can see, and the metric in which limits are set. - Evidence: optical (PCM) rather than electron (TEM) microscopy used routinely “mainly because of cost and availability”, and therefore in the dose–response epidemiology and in regulations; TEM detects the smallest fibres; at the Dutch limit, TEM counts about twice as many fibres as PCM (20,000 vs 10,000 fibres/m³) (p. 725). - Strength: suggestive to moderate. The mechanism is clearly stated in one paragraph. The annex does not show that it changed any limit; the much lower recommended Dutch limits (roughly 10–50 times below the current limit) reflect a risk-based derivation, not instrument choice.

5. “Safe” or “background” levels are provisional judgments. They can erode, and can cancel a product’s claimed benefit. - Evidence: p. 718 (POPs abolishing fish benefits in rodents); p. 720 (benzene below 1 ppm); p. 726 (WHO: no asbestos exposure free of cancer risk); p. 717 (editors’ synthesis). - Strength: suggestive to moderate. Strong for asbestos and benzene; for fish it rests on two rodent studies by the author.

6. Benefit claims deserve the same evidential scrutiny as risk claims. Long-accepted benefits can rest on small, unrepresentative studies and shrink under rigorous testing, which shifts the burden of justification. - Evidence: pp. 732–733 (7 chicks for 9 days; the Perdue crossover study; USDA survey; Graham reanalysis). - Strength: moderate. The key benefit study was well designed and industry-originated. The “net loss” conclusion is the author’s own reanalysis, though it assumed a significant performance gain and still found costs exceeded it (p. 733). [External — verify: later USDA work suggests small positive benefits, which still supports “smaller than claimed”.]

7. Practices can outlive the rationale that justified them. - Evidence: growth promoters as a 1940s substitute for costly vitamin B12, persisting after B12 synthesis and husbandry advances (p. 733). - Strength: suggestive. Historically plausible; the author frames it with “may”.

8. Systems optimised for their operators’ metrics can degrade properties that users rely on, and even when the change is measured its consequences for users may go unstudied. - Evidence: aquafeed optimised for fish growth and health; the EPA+DHA to n-6 ratio fell about 50% in 2005–09 (measured by NIFES), with health consequences “not researched” (p. 718). - Strength: suggestive. The compositional change is data; the health consequence is inferred.

9. Substitutes can bring new problems, but well-chosen substitutes can outperform the incumbent. Both outcomes need assessing before and after a switch. - Evidence: new chemicals from vegetable feeds (p. 718); ozone-substitute gases warming the climate (p. 728); asbestos substitutes with superior properties or lower cost (p. 725). - Strength: moderate. Documented in both directions, though briefly.

10. Safer alternatives often cost little relative to the whole system, and some of their extra cost is recovered through avoided downstream costs. Incumbents’ claims that alternatives are prohibitively expensive should be tested.

    • Evidence: USD 65 extra per town-house roof in Thailand, 0.5% of overall cost (p. 724, unsourced); 10–15% manufacturing premium for fibre-cement panels “to some degree offset” by avoided hygiene, hazard, removal and disposal costs (World Bank, p. 725).
    • Strength: moderate to strong for building products (World Bank quotation; Thai figure unsourced). Generalising beyond them is inference.

11. Restricting a hazard in some jurisdictions can push it into others, and exporting interests may block international information-sharing mechanisms.

    • Evidence: bans in 50+ countries “offset” (p. 724); Canada’s exports and its blocking of the Rotterdam listing (pp. 725–726); Russian pressure on Thailand (p. 724); growth promoters spreading with industrial production (p. 731).
    • Strength: strong for exporter obstruction of information-sharing (specific, dated asbestos events). Moderate for displacement itself: the annex shows bans “offset” by growth elsewhere and continued exports, but not that the bans caused the growth. Moderate for antimicrobials, where the spread is described but not tied to restrictions elsewhere.

12. Where harm accumulates in a shared, pooled resource, each local use imposes system-wide costs. Assessment frames that judge each use by its immediate target miss this.

    • Evidence: “one reservoir”; horizontal gene transfer warned about in the early 1970s but ignored (p. 732); O’Brien’s “anywhere … anywhere else” (p. 731).
    • Strength: moderate to strong. Established microbiology, though the annex does not quantify the share of human burden due to animal use.

13. How experts and regulators frame a question (which endpoints, units and uses count) can exclude known mechanisms for decades.

    • Evidence: the “old notion” of clinical importance (p. 732); the WTO’s “restrictive notion of risk assessment” (p. 737).
    • Strength: moderate.

14. Risk assessment should include real-world conditions of use, including misuse and poor compliance.

    • Evidence: the WTO Appellate Body on “misuse and abuse in the administration of hormones” (p. 737).
    • Strength: moderate. A legal finding reported by one party’s counsel; the principle itself is clear.

15. Whether a precautionary measure survives adjudication depends on procedural rules: expert independence, burden of proof, whether reviewers pick the “best science” or check the regulator’s reasoning, and how “insufficient evidence” is defined.

    • Evidence: pp. 736–737.
    • Strength: moderate. The legal record is quoted, but the summary is partisan, the AB “could not complete the analysis” (p. 737), and [External — verify] the dispute ended in negotiation, not a merits ruling.

16. Legacy stocks and long latencies mean harm continues long after a ban and relief arrives decades later. Recovery depends on sustained compliance.

    • Evidence: asbestos in place and Peto’s 50-year projection (pp. 724–725); the Netherlands two decades after its ban (p. 725); ozone recovery mid-century “assuming global compliance” (p. 728); DES in the third generation (p. 729).
    • Strength: strong.

17. Information asymmetry protects incumbents. Use and benefit data often sit with industry and may need compulsion to release. Independent monitoring after a policy change can supply decisive evidence.

    • Evidence: US “compelled the release of industry data” (p. 731); resistance declines after avoparcin removal (p. 732); Danish industry experience, which the annex mentions as “careful documentation” but does not cite or describe (p. 732).
    • Strength: moderate. The 80% statistic is framed loosely (see Limitations).

18. Early-acting jurisdictions serve as natural experiments that can test predictions of economic harm from regulation.

    • Evidence: Denmark and Netherlands bans with no productivity loss and improved husbandry (p. 731).
    • Strength: moderate. The author’s characterisation; details (e.g. weaner-pig effects) are contested.

19. Regulatory pressure can prompt process and product innovation that offsets expected costs.

    • Evidence: husbandry improvements (p. 731); fibre-cement products with superior properties (p. 725). Koppe’s claim that “denial” blocked curative innovation (p. 727) is asserted only.
    • Strength: suggestive. Illustrations only.

20. Transnational coalitions, media exposure, elections, courts, sub-national governments and individual officials can shift entrenched positions, often locally first.

    • Evidence: p. 726 (Osasco and the South American bans; the Asian network; Brazil’s states and Supreme Court; Giannasi; CBC; the Quebec election; the WHO rebuke).
    • Strength: moderate. Documented events, written by participants, with no analysis of how much each factor contributed.

21. Once evidence is strong, the question shifts from “why were early warnings missed?” to “why does the practice persist?”. Understanding the motives and structure of opposition becomes the analytical task.

    • Evidence: pp. 731, 733.
    • Strength: asserted to moderate. “Overwhelming evidence” is the author’s judgment, and the called-for analysis of motives is not done here.

22. Chemically or technically distinct agents can act on the same biological targets, so their effects may combine, and agent-by-agent assessment would miss this.

    • Evidence: the same 14 genes altered in the same direction in a PCB-exposed group and in a separate DDT/DDE/HCB-exposed group; Koppe reads this as “the additive effect” (p. 727).
    • Strength: suggestive. A single study, loosely described (gene expression called “epigenetic”). Shared targets are shown; additivity is the author’s inference, since the groups were exposed separately.

23. Laboratory evidence may be discounted until human evidence arrives, producing lags of decades.

    • Evidence: formaldehyde, 1989 lab results to the 2009 IARC human-evidence conclusion (p. 735).
    • Strength: suggestive. One instance, told by the lab concerned, whose own methods have been contested.

Limitations, contestation and bias check#

Genre and purpose. The editors describe these as “brief” updates, and full reviews were “a step too far” (p. 717). They should not be judged as systematic reviews. But the editors use them to support a general thesis (p. 717), and at that point their selectivity matters.

Selection and composition. - Nine of fourteen cases are updated. There are no industry or dissenting voices and no panels; the one regulator voice (the European Commission’s lawyer) is a party to the dispute he describes. - Most authors are revisiting their own cases or promoting their own laboratory’s findings. - The one legal summary is written by counsel for one of the parties (the EC), who characterises the outcome as a victory. - The one recovery story (ozone) is presented without comment and without its implications. - This composition inclines the annex toward confirming the “harm expands” narrative.

Self-reference and unpublished evidence. - Ruzzin: 3 of 14 references are his own, and they carry the causal core (p. 719). - Infante: 7 of about 38 references are his, including two manuscripts “submitted for publication” that carry the Pliofilm claim (pp. 721–722). - Silbergeld: 4 of 13 references, including the economic reanalysis on which “net loss” rests (pp. 733–734). - Belpoggi cites none of the MTBE bioassays behind her central claim (p. 735). - No declarations of interest appear anywhere.

Evidence thinness. - PCBs (p. 727): 3 references; key claims uncited; inference by analogy from famine and from a rat growth-retardation model that the annex does not say involved PCBs. - MTBE (p. 735): the central claim is uncited, supported by ecological trends and a metabolite argument. - Fish (p. 718): rodent data plus heterogeneous human associations; the nutritional-impact claim is admittedly unresearched. - Unsourced key numbers: the asbestos “5–10 million” eventual deaths, the WHO death and exposure estimates (attributed to WHO without a reference), the consumption figures and the Thai USD 65 roof figure (p. 724); the Dutch diabetes prevalence and EUR 4.5 bn cost (p. 727); MRSA as the US’s “single most important cause” of infectious morbidity and mortality (p. 731).

Counter-evidence or counter-arguments not engaged (partly visible in the annex itself): - The Science exchange challenging low-dose benzene blood toxicity (Lan et al., 2006, “Response to Lamm and Grunwald”, p. 722) is cited but not discussed. - Other Pliofilm reanalyses (Crump, 1996; Paxton, 1996, p. 722) are listed but not discussed. - IARC’s narrower 2009 conclusion on benzene is disputed only by title (Infante, 2011, p. 722). - For fish, there is no benefit–risk weighing. [External — verify: FAO/WHO 2010.] - For MTBE, IARC’s 1999 Group 3 classification and the contested Ramazzini pathology go unmentioned. [External — verify.] - For antimicrobials: the EU-wide 2006 growth-promoter ban and FDA’s 2012 guidance are not mentioned [External — verify]; the Danish transition problems in weaner pigs are not acknowledged [External — verify]; and “no dispute” is offered as though it were confirmation (p. 733).

Internal errors and inconsistencies (verifiable from the page). - The editors’ DES example is loosely worded and can be read as conflating vaginal-cancer survivors with the broader group of DES daughters at raised breast cancer risk (p. 717 vs. p. 729). - The hormones footnote gives AB-2006-5/6 while the heading, and the footnote’s own next sentence, say AB-2008 (p. 736). - Silbergeld calls her two key benefit studies “two recent industry studies”, but the second is a USDA survey analysis; the Danish “careful documentation” is not cited (p. 732). - “Bridgess” (p. 736) and “MBTE” (p. 735) are misspellings of the Vol. 1 attributions. - The O’Brien quotation is duplicated (p. 731). - The UK compensation reference muddles court levels (p. 726). - Koppe’s Netherlands diabetes sentence is garbled (p. 727). - Infante’s in-text “Paxton et al., 1994” does not match the reference list’s “Paxton MB., 1996” (pp. 721–722). - The Valvi reference is dated 2011 but cited to EHP vol. 120 [External — verify: vol. 120 is 2012; probably online-first in 2011, so a trivial discrepancy] (p. 727). - Possible transcription error in the DES control count (1,027) [External — verify: 1,927]. - The BPA date of “1930” [External — verify: 1936].

Framing effects. - The editors’ thesis (p. 717) is built partly on hindsight: cases were chosen for Vol. 1 because harm materialised. “Harm expands” is then observed in a sample conditioned on harm. - The annex does not ask how often early warnings were not borne out, or how often intensive study of a suspect agent generates spurious associations (e.g. the annex reports 12 raised outcomes in DES daughters without saying how many were examined, p. 729). - The language of “denial”, “intransigence” and “victory” signals advocacy, not analysis.

Fairness in the other direction. - The advocacy is transparent, not disguised: titles such as “the campaign for a global ban” and “personal summary” declare it. - Much of the empirical content is sourced to high-quality peer-reviewed work: the NCI DES cohorts, Science papers on low-dose benzene, the industry-run Perdue trial, World Bank and WHO documents. - Several warnings in the annex were strengthened by later events, all [External — verify]: - BPA restrictions and a drastically lowered TDI; - IARC Group 1 for PCBs; - Canada’s asbestos ban, the US EPA chrysotile ban and the lower EU occupational limit; - US and EU action on growth promoters; - lower EU occupational limits for benzene. - Some warnings remain contested or were weakened [External — verify]: MTBE carcinogenicity; the breadth of benzene’s cancer spectrum; the fish–diabetes link. - A fair reading: the annex is right on the direction of travel for several agents, while individual claims range from strong (DES, asbestos politics) to asserted (PCBs, MTBE).

What the annex cannot tell us. None of the pieces analyses why institutions behaved as they did in any depth. Silbergeld explicitly calls for that analysis and does not supply it (p. 731). The mechanisms above are inferred from outcomes and from partisan narration, not from documentary evidence of decision-making.

Notable quotes#

1. “their capacity to harm expands over time” (editors, p. 717)
2. “background levels of POPs, which many people consider to be at safe levels, can fully abolish the potential health benefits of fish nutrients” (Ruzzin, p. 718)
3. “quantitative cancer risk assessments have underestimated cancer risk resulting from environmental benzene exposure levels” (Infante, p. 721)
4. “Bans in over 50 countries have been offset by dramatically increasing use in China, India and other emerging economies.” (Castleman and Gee, p. 724)
5. “mainly because of cost and availability considerations, phase‑contrast microscopy has been routinely used” (Castleman and Gee, p. 725)
6. “Denying toxic effects of PCBs has not only made more people ill, but has also hampered the innovative development to cure the epigenetic changes.” (Koppe, p. 727)
7. “the long‑term consequences we wrote about in 2000 were merely the tip of the iceberg” (Swan, p. 729)
8. “a case study of intransigence in the face of overwhelming evidence” (Silbergeld, p. 731)
9. “persistence in a clearly dangerous practice in the absence of benefit” (Silbergeld, p. 733)
10. “somewhat peremptorily decided what it considered to be the best science” (WTO Appellate Body, para 612, as quoted by Christoforou, p. 737)

Open questions#

1. Is “harm expands over time” a general property of hazardous agents, or an artefact of selection and research intensity? What would a non-selected sample of early warnings, including false alarms, show?
2. Were Infante’s Pliofilm “missing deaths” analyses published and accepted? Did regulators adopt the low-dose metabolism findings or the broader cancer spectrum (IARC 2017; EU limits)?
3. Has later human evidence supported POP-driven metabolic harm from farmed fish, or the claim that fish benefits are overestimated? Did EU regulation of PFAS, PBDEs and other pollutants in seafood follow, and did dietary advice change?
4. What happened to global asbestos consumption, the Rotterdam listing, and bans in India, China and Russia after 2013? Did the European Parliament’s 2028 removal target become binding?
5. Has the PCB–type 2 diabetes link held in longitudinal studies, and is there any evidence behind the “denial hampers curative innovation” claim?
6. Did third-generation DES effects materialise? How did BPA regulation evolve, and how long was the lag from first knowledge of its hormonal activity to restriction?
7. What are the best current estimates of growth-promoter productivity effects? Did US policy after 2013 and EU Regulation 2019/6 reduce use? How much did use grow in Brazil, China and India? What share of human resistance is attributable to animal use?
8. What is MTBE’s carcinogenicity status after 2013 (IARC 2025), and how was the dispute over Ramazzini data resolved?
9. Why did the hormones dispute end in a negotiated market-access deal, not a ruling? Have the Appellate Body’s standards (expert independence, misuse, “critical mass”, international standards) shaped later disputes, especially now that the Appellate Body itself has been non-functional since 2019 [External — verify]?
10. Why were five Vol. 1 cases (e.g. BSE, TBT, radiation) not updated, and would their updates have supported or complicated the editors’ thesis?
11. The ozone case is the annex’s clearest success. What features (a limited number of producers, available substitutes, a visible signal, a treaty with trade measures) distinguish it from asbestos and antimicrobials, where displacement and blocking prevailed?

Audit log#

Independent audit against the full text extract (PDF 719–739) with spot checks of the PDF (no images on any annex page; footnote on p. 736 and key numbers confirmed). Quotations checked verbatim; one misquote found. External flags left as flags (not web-checked in this pass).

1. How-this-was-read: corrected “about 40%” of the extract being references to “about a quarter” (measured ~24% by characters).
2. Section overview: “no industry or regulator responses” corrected, since the European Commission’s lawyer (a regulator and party) is a contributor; same fix in Limitations.
3. Infante standpoint: softened “paper that first reported leukaemia” to what the annex shows (the 1977 paper heads the Pliofilm citations).
4. Editors’ intro: softened the DES “conflation” to a loosely worded, ambiguous sentence; changed “PCBs and fish support it only by assertion” to distinguish PCBs (assertion) from fish (author’s rodent studies plus mixed human data).
5. Fish: fixed misquote (“introduce” to “introduces new hazardous chemicals that may threaten human health”).
6. Fish: corrected “nobody was assigned to track it”; the compositional change was measured (NIFES); only its health impact was unresearched. Carried through to Mechanism 8, Insight 8 and the digest.
7. Fish external note: Reg. (EU) 1259/2011 predates the annex, so it cannot have “later” closed the gap; reworded.
8. Benzene: added the unsourced “most toxic aromatic hydrocarbon” claim; restored the hedge “may be a reflection of” on the AML/cytopenia explanation.
9. Benzene appraisal: replaced “and he knew it” with a neutral statement; added that childhood-leukaemia studies rest on exposure proxies, not measured benzene.
10. Asbestos: noted the WHO death/exposure figures and consumption data carry no specific reference; separated “thousands more” from the WHO attribution.
11. Asbestos: corrected the Thai figure: USD 65 per town-house roof adds 0.5% to the overall cost, not “0.5% more for a roof” (fixed in section notes, Mechanism 6, Insight 10 and digest); flagged the figure as unsourced.
12. Asbestos: attributed “superior impact resistance” specifically to polypropylene-cellulose-cement roofing, as the World Bank quote does.
13. Asbestos/measurement: added the annex’s TEM/PCM conversion (20,000 vs 10,000 f/m³) and noted the recommended Dutch limits reflect risk-based derivation, not instrument choice; Insight 4 downgraded to suggestive-to-moderate (digest too).
14. Asbestos: added external-verify flags on the 1999 WTO date (consultations May 1998) and the 4 Sept 2012 date (election day); timeline now marks Ottawa’s Rotterdam shift as undated in the annex.
15. Asbestos Mechanism 1: “roughly an order of magnitude” changed to “roughly 10–50 times” below the current Dutch limit.
16. PCBs: noted the one diabetes citation (Lee et al., 2006) concerns POPs generally; the growth-retardation claim is effectively uncited; Mitra et al. studied Slovakian girls.
17. PCBs: noted that same-direction gene changes in separately exposed groups show shared targets, not additivity; Insight 22 and the Mechanism 2 item reworded and labelled as inference.
18. Ozone: added omitted details (early-October peak, hole larger than Antarctica, latitude gradient); softened “only update … working” and “explicitly flags regrettable substitution” to what the text says.
19. DES: restored the “for most outcomes” qualifier on the VEC gradient; corrected the claim that twelve outcomes were “examined” (the annex reports twelve raised outcomes, not the number tested), here and in Limitations.
20. Antimicrobials: added omitted claim that “overall drug uses have increased substantially” and noted it is inferred from production rankings.
21. Antimicrobials: corrected “studies … show” to the annex’s “studies have been conducted … on the prevalence”; flagged the EU livestock-MRSA statement as uncited.
22. Antimicrobials: flagged the internal inconsistency that the “two recent industry studies” include a USDA survey, and that the Danish documentation is uncited (section notes, Mechanism 12, Insight 17, Limitations, digest).
23. Antimicrobials: added the key nuance that Graham et al. assumed a significant performance benefit and still found a net loss (section notes, Insight 6, digest); added Perdue study location and description.
24. Antimicrobials: added Silbergeld’s concession that growth promoters may once have conferred real benefits (section notes, digest).
25. Antimicrobials appraisal: replaced “appears overstated” for the MRSA claim with external context; tagged the EU 2006 growth-promoter ban as external.
26. Mechanism 10: fixed garbled sentence (what was “apparently ignored” was horizontal gene transfer, not the clinical-importance model).
27. MTBE: added Belpoggi’s pairing of female-rodent tumours with a female-only human trend; tagged the “reframing” and endogenous-formaldehyde points as external.
28. Hormones: recorded that the footnote is internally inconsistent and that paragraph numbers refer to AB-2008-6; added the “clear roadmap” claim and the admission that the AB could not complete the analysis.
29. Mechanisms 6 and 15: labelled the incumbents’-cost-claims point and the producer-benefit reasoning as the note-taker’s inference.
30. Insight 11: split strength: strong for exporter obstruction, moderate for displacement (causation not shown); digest item 8 adjusted.
31. Insight 15: tagged “ended in negotiation” as external and added the AB’s incomplete analysis.
32. Timeline: DES cohort dates corrected to 2008–2011.
33. Recommendations: added that the ozone and benzene pieces make no explicit recommendations.
34. Limitations: widened the list of unsourced key numbers; softened the Valvi “error” to a trivial, externally flagged discrepancy.
35. Digest: fixed benzene hedges (“appears”, “likely”), asbestos numbers, PCB “blocked cures” (to “hampered development of cures”), “ignored” (to “apparently ignored”), ozone “one success story” (to “clearest”), and rewrote the hormones bullet to use the annex’s wording and include the incomplete analysis and the sanctions outcome.