LL2-07 — Ch7 Tobacco industry manipulation of research#

Late lessons from early warnings: science, precaution, innovation (EEA Report No 1/2013), Part A “Lessons from health hazards”. Report pp. 151–178 (PDF pp. 153–180). Chapter text runs pp. 151–171; references pp. 172–178.

Reading notes: I read the full text extract page by page, through the final page marker (PDF 180 / p. 178). I checked the opening box (p. 151), Box 7.1 (p. 155), Table 7.1 (p. 161), Panel 7.1 (pp. 164–165), Box 7.2 (p. 168) and Table 7.2 (p. 171) against rendered PDF page images. The extract is faithful for these; the only garbling is tab-spacing in bulleted lists. The photographs on pp. 160 and 170 carry no captions beyond stock-photo credits.

Where I add context from outside the report, I mark it [External]. Items marked “verified” were checked this session against primary records (Europe PMC abstracts, the Federal Register API). Items marked “unverified” come from general background knowledge and should be checked before they are cited.

Authors and standpoint#

Lead author: Lisa A. Bero (sole author). The chapter does not give her institutional affiliation. Its footnote (p. 151) lists funding from the American Cancer Society and the University of California Tobacco-Related Disease Research Program. It thanks collaborators on the cited studies (Bryan-Jones, Glantz, Hong, Mangurian, Montini, Schotland) and, separately, others for “useful input to this manuscript” (including David Rosner). The footnote says “Authors would like to thank” (plural) although the chapter has one author. - [External, unverified] At the time Bero was a professor at the University of California, San Francisco (clinical pharmacy and health policy). She co-authored The Cigarette Papers (Glantz et al., 1996, cited here) and is a leading empirical researcher on conflicts of interest and research integrity. She has held senior roles in the Cochrane Collaboration. - The chapter says it is “an expansion of an earlier article” (Bero, 2005, Public Health Reports) (p. 152). Most of its evidence dates from 1993 to 2006, and some phrasing is dated. For example, the 2005 special issue of the International Journal of Occupational and Environmental Health is called “a recent issue”, and the eight strategies are called “the strategies outlined in this article” (p. 169), wording carried over from the journal article. - It is largely a synthesis of the author’s own research programme. About 34 of the cited works include Bero as an author (Bero & Glantz 1993; Bero et al. 1994, 1995; Barnes & Bero 1996, 1997, 1998; Hong & Bero 2002; Schotland & Bero 2002; Misakian & Bero 1998; Kennedy & Bero 1999; Mangurian & Bero 2000; Bero et al. 2001; Bryan-Jones & Bero 2003; Lexchin et al. 2003; and others). This is a strength: the chapter reports primary, peer-reviewed empirical work first-hand. It is also a limitation: the chapter is not an independent review of that work. - Evident stance: public-health and research-integrity oriented, and openly critical of the tobacco industry. It ends in explicit normative recommendations: full disclosure, with bans on tobacco-industry research funding “warranted” (p. 170). The tone is mostly measured and empirical, and inferences about motive are often hedged (“It is therefore possible that…”, “may have also been funding…”, p. 157).

Other components, and who wrote them - Opening summary box (p. 151). Unsigned. It appears to be the EEA editors’ framing, in the format used to open other chapters. It says the chapter “differs in some ways from the others”, because it concentrates on “argumentation” rather than on the history of scientific knowledge. It also adds a generalisation, with no source: “other industries are drawing inspiration from tobacco company strategies, seeking to maintain doubt about harm in order to keep hazardous products in the marketplace” (p. 151). The chapter body is not uniformly more cautious. It too asserts that the industry’s efforts “are shared by other corporate interests” (p. 152) and that its methods “are similar to those of other corporate interests” (p. 167), but it also says more must be learned before drawing conclusions about similarities (p. 168). The box’s “drawing inspiration” (deliberate copying) goes further than any of these. - Panel 7.1, “Shaping risk assessment in the US and the EU: the role of the tobacco industry” (pp. 164–165). By Katherine Smith, Anna Gilmore and Gary Fooks. No affiliations are given. [External, unverified] At the time they were tobacco-control researchers at the University of Bath. The panel rests mainly on their own analysis of British American Tobacco (BAT) documents (Smith et al., 2010, PLoS Medicine; verified via Europe PMC). It is not a dissent. It shares the chapter’s critical stance and extends it to EU regulatory reform. It closes with a balancing sentence: corporate input is “of course legitimate” if it is transparent, “not excessive”, and does not compromise public welfare (p. 165). - Box 7.2 (p. 168). A verbatim quotation from the WHO Committee of Experts on Tobacco Industry Documents (CETID, 2000). This is an official WHO-commissioned inquiry, not the chapter author’s voice. - Table 7.2, “Key dates relating to knowledge of harm from active and second-hand smoke” (p. 171). Source line: “EEA, based on Keating, 2009 and Ong and Glantz, 2000”. It was compiled by the EEA, not by Bero, and relies heavily on a biography of Richard Doll (Keating 2009). It contains apparent errors (see Limitations). - No industry or regulator response panel. Unlike some chapters, there is no rejoinder from the tobacco industry or from scientists it funded. Readers therefore get no counter-account of specific disputed episodes. [External, from an earlier Europe PMC check; not re-verified in this audit] For example, the Yano (2005) reanalysis was published with a discussion section (pp. 233–5 of that issue) that the chapter does not mention.

Section-by-section notes#

Opening box (p. 151; likely EEA editorial)#

• Subject: second-hand (“passive”, “environmental”) tobacco smoke (ETS/SHS). Those affected: “the partners and children of smokers, and the bartenders and other workers who have to work in smoky environments”.
• Focus: strategies to “deny, downplay, distort and dismiss the growing evidence” of harm. The chapter deliberately does not retell the history of tobacco science; Table 7.2 stands in for it.
• Guiding questions: who generated and financed the refuting science, why, and with what “science and information, tools and assumptions”.
• Flags: litigation-released documents revealed the pattern, but “this insight is not available for most corporate sectors”. The box trails “full disclosure” and Panel 7.1.

7.1 Introduction (pp. 152–153)#

• Core claim: “The primary motivation of the tobacco industry has been to generate controversy about the health risks of its products” (p. 152).
• Four summary strategies (later expanded to eight in Box 7.1): fund and publish supportive research; suppress and criticise unsupportive research; change research standards; disseminate interpretations to the lay press and policymakers.
• Constancy: the strategies “remained remarkably constant” from the early 1950s (active smoking) to the 1990s (SHS). “Tobacco industry lawyers and executives, rather than scientists, have controlled the design, conduct and dissemination of this research” (p. 152).
• Early prescription: when risk data appear controversial, “users of the data should investigate the sources of controversy”. This requires interest-group involvement to be “transparent and fully disclosed” (p. 152).
• Theory (STS / policy studies):
• evidence is socially constructed (Krimsky); framing and language shape risk communication (Nelkin);
• uncertainty allows “a wide range of interpretation of the same data”;
• “Since data do not ‘speak for themselves’ interest groups can play a critical role…” (p. 152).
• Interest groups defined neutrally (Lowi; Walker; Truman). Public-health groups are “likely to” emphasise harm and industry groups “likely to” minimise it (an expectation, not a finding). Both aim to influence policy and litigation (Jasanoff). The symmetry is conceded here but not pursued later.
• “History repeating itself” from the 1950s. Other corporate groups “appear to use similar tactics” (Special Issue 2005; White et al. 2009).
• Acknowledged selection effects (pp. 152–153): the strategies are costly, so industry “might therefore be expected to dominate examples”. Public-health groups “are less likely to command such resources”. Industry examples also predominate because of documents released in “discovery” (asbestos, tobacco).

7.2 Scientific knowledge of SHS hazards (p. 153)#

• Definition: ETS is “thousands of gases and fine particles” from burning tobacco and exhaled smoke.
• Evidence chronology:
• 1970s–80s: data begin appearing.
• 1981: “Seminal epidemiological studies… demonstrated” an association with lung cancer (only Hirayama 1981 is cited).
• 1986: the US Surgeon General and NRC conclude SHS is “a cause of disease”.
• 1988–1998: IARC European study reports “a 16 % increase… for non-smoking spouses… and a 17 % increase” for workplace exposure. No confidence intervals are given.
• 1992: the US EPA classifies ETS as a “Group A human carcinogen”.
• EPA nuance:
• The industry criticised study selection, statistics, design, bias and confounding control, and exposure measurement (Bero & Glantz 1993).
• “The EPA revised the report in response to valid criticisms”. The report “was improved but the sheer volume of tobacco industry comments… probably delayed its release”.
• Some criticism was valid. The chapter says the volume of comments “probably” delayed release. On p. 166 the same delay is stated without the hedge: development “was considerably delayed by the tobacco industry’s criticisms”.
• Court: “Although the science was valid, the tobacco industry successfully attacked the US EPA risk assessment in court on procedural grounds (Flue-Cured… 1998)”. There were similar procedural objections to Australia’s NHMRC 1997 report. The main text does not say what happened to the ruling afterwards [External, unverified: it was vacated on appeal in 2002]; the panel refers only to a “cloud… until 2002” (p. 164).
• California EPA:
• 1997: broader scope, and it addressed the industry’s criticisms of the US EPA. SHS is “a toxic air contaminant” that “has a major impact on public health”.
• 2005 update: causal for developmental effects (foetal growth, SIDS, pre-term delivery); respiratory effects (asthma, childhood symptoms, ear infections); cancers (lung, nasal sinus, “breast cancer in younger, premenopausal women”); and cardiovascular effects.
• Motive: SHS evidence was “clearly a threat… as early as the 1970s”, because restrictions “could lead to reduced daily consumption of cigarettes and a decline in sales”. The industry “responded with its own science” (p. 153).

7.3 Strategies to subvert scientific knowledge (pp. 153–154)#

• Engine: “Policymaking is facilitated by consensus… Scientific research, on the other hand, is characterised by uncertainty”. Controversy “is likely to slow or prevent regulation”. Debate over data and methods hinders risk assessment (Stayner 1999) (pp. 153–154).
• Source base: the 1998 MSA documents show decades of motivation to generate controversy, and concern for “maintaining its credibility as it manipulated research” (p. 154).
• Key internal quotations (p. 154):
• Roper Organization for the Tobacco Institute (1978): the “strategic and long-run antidote” is “developing and widely publicizing clear-cut, credible, medical evidence that passive smoking is not harmful”.
• UK industry meeting, London, 17 February 1988 (notes by Boyse, a BAT document): Philip Morris proposed, “in key countries” within every major international area, teams of scientists “organized by one national coordinating scientist and American lawyers”, “to review scientific literature or carry out work on ETS to keep the controversy alive” (the chapter bolds “keep the controversy alive”: “emphasis added”).
• Consultant teams worldwide had the “main goal of stimulating controversy” (five document studies cited).
• Sponsorship and outcome:
• Sponsorship correlates with favourable outcomes, but “there is no consistent association between industry sponsorship and methodological quality” (Lexchin et al. 2003).
• Outcomes are shaped instead by “the framing or social construction of the research question; the conduct of the study; the publication (or not)”.
• The industry manipulated all three: through agenda-setting funding, lawyer and executive involvement in design and conduct, and sponsored publication combined with suppression (p. 154).

Box 7.1: eight strategies (p. 155)#

1. Fund supportive research.
2. Hide industry involvement.
3. Publish supportive research.
4. Suppress unsupportive research.
5. Criticise unsupportive research.
6. Change scientific standards.
7. Disseminate to the lay press.
8. Disseminate directly to policymakers.

The same list is repeated on p. 169 as a template also used by chemical companies.

7.3.1 Strategy 1: fund research (pp. 154–157)#

• Functions (p. 155):
• agenda control;
• refutation data;
• material for press and policymakers;
• PR as “a philanthropic body”;
• credibility. A Philip Morris Worldwide Scientific Affairs funding criterion was “whether the research would enhance the credibility of the company” (Malone & Bero 2003).
• Channels and control: the Tobacco Institute, internal research, consultants and sponsored organisations. Lawyers and executives chose projects. “Most of the research did not undergo any form of independent scientific peer review but was funded on the basis of its potential to protect the interests of the companies” (p. 155).
• Lawyers (pp. 155–156):
• Documents leaked through whistle-blowers in the mid-1990s. Availability rose “exponentially” after the 1998 Minnesota/Blue Cross settlement.
• The MSA (a settlement “between the attorneys general of 46 states and” Brown & Williamson/BAT, Lorillard, Philip Morris, RJ Reynolds, CTR and the Tobacco Institute) released “millions” more documents.
• Lawyer involvement “protected their research activities from public discovery and kept their lawyers informed about science relevant to litigation”.
• Law firms (Covington & Burling; Jacob & Medinger) funded projects directly: literature reviews, and confounder research on genetics, stress and low-protein diets, “deflecting attention away from tobacco”.
• Consultants were paid to prepare Congressional testimony.
• Brown & Williamson’s legal department controlled dissemination of internal science and screened reports for legal privilege.
• Wells memo (17 February 1986; J. K. Wells, B&W corporate counsel): “The only way BAT [British American Tobacco] can avoid having information useful to plaintiff found at B&W is to obtain good legal counsel and cease producing information in Canada, Germany, Brazil and other places that is helpful to plaintiffs” (p. 156).
• Public claims vs practice: public statements said the companies funded “objective research”, but lawyer involvement “served to control the scientific debate… and protect from discovery scientific documents that were potentially damaging” (p. 156).
• CTR (formerly TIRC, 1954), p. 156:
• Publicly, it was to fund independent research. Internally it “was actually formed for public relations purposes, to convince the public that the hazards of smoking had not been proven” (Glantz et al. 1996).
• Its Scientific Advisory Board of “well respected researchers” did not review all projects. From 1966, lawyers “became directly responsible for many of the funding decisions”.
• 1972–91: “at least USD 14 636 918 in special project funding”. Lawyers also designed the studies and disseminated the results.
• Pepples memo (4 April 1978, general counsel to CEO), p. 156: “Originally, CTR was organized as a public relations effort… also discharged a legal responsibility… another political need for research”. It described rapid response to “the propaganda of the avid anti-smoking groups” and “special projects designed to find scientists and medical doctors who might serve as industry witnesses”.
• CIAR (1988), pp. 156–157:
• Founded by Philip Morris, RJ Reynolds and Lorillard; Svenska Tobaks joined in 1994.
• Stated mission: “quality, objective research in indoor air issues including environmental tobacco smoke”. The ETS wording was dropped in 1992, followed by an agenda shift “that would prevent it investigating the health effects of second-hand smoke”.
• 1989–93 funding: USD 11 209 388 “peer-reviewed” (Science Advisory Board) and USD 4 022 723 “special-reviewed” (board of tobacco executives).
• 70% of peer-reviewed projects examined pollutants other than tobacco smoke, read as credibility, publicity and diversion.
• Almost two-thirds of special-reviewed projects concerned SHS, mostly exposure rather than health. (Both the 70% and the two-thirds are shares of projects, not of funding.) The author calls it “possible” that these were meant to support the claim that exposures are too low to harm.
• Six CIAR-funded investigators testified at hearings, and “All… supported the tobacco industry position”. Data went to OSHA hearings (two projects) and to a Congressional hearing on an aircraft smoking ban (one).
• Fabrication: Turner et al. (1992) concluded that “with good building ventilation, clean air could be maintained with moderate amounts of smoking”. It was used against restrictions, and a Congressional subcommittee found its data “altered and fabricated”. An earlier study was compromised because the Tobacco Institute chose the measurement sites.
• Reinvention: CIAR was disbanded under the MSA (1998). The Philip Morris External Research Program (PMERP, 2000) copied its structure. Three of its six advisory board members had CIAR ties, and most named reviewers had prior industry ties (Hirschhorn et al. 2001).
• International (p. 157): Covington & Burling consultant networks in Latin America and Asia. In Germany, companies and the German Association of the Cigarette Industry funded research through foundations (Grüning et al. 2006).

7.3.2 Strategy 2: hide involvement (pp. 157–158)#

• INBIFO:
• 1968: Philip Morris plans an SHS animal-research facility. 1970: buys INBIFO (Germany).
• Coordinator Ragnar Rylander reported to US executives, “who would then decide whether to disseminate… or keep it secret”.
• Unpublished: sidestream smoke “more toxic than ‘mainstream smoke’”. Published: a green-tea and lung-cancer association (Tewes et al. 1990), a distraction.
• Rylander had “long-standing and secret links” and “conferred a false sense of credibility”. The University of Geneva Fact Finding Commission (2004) found him “a sponsored agent of the tobacco industry, rather than… an independent researcher”. He took no SHS initiatives “without first consulting extensively with his contacts within the tobacco industry” (p. 158).
• Japanese Spousal Smoking Study (p. 158):
• Hirayama (1981) was voted the most influential SHS paper and was the most cited in regulatory hearings. The industry attacked it for misclassification bias.
• Per Hong & Bero (2002): the named investigators were Japanese, but Covington & Burling managed the project, an industry scientist supervised it, and an industry consultant reviewed the design and interpreted the data. None of these were to be co-authors.
• The funders considered CIAR as “a cover” but funded directly through three companies. Progress reports were on law-firm stationery.
• The consultant (Lee 1995) was sole author, with an acknowledgement of “financial support from several companies of the tobacco industry”, which “tells the reader little”.
• Original investigator E. Yano was unaware of the publication. He kept the data and showed that Lee “excluded data that did not support misclassification bias”. The full data change the conclusion: “After 10 years, the scientific community was able to obtain data that had been suppressed”.

7.3.3 Strategy 3: publish (pp. 159–161)#

• “Research has little impact unless it can be cited”. Vehicles: symposia, books, articles and letters, then cited “in scientific and policy arenas” (p. 159).
• Precedent: the chapter notes that pharmaceutical-industry symposium reports also contain “poor quality and unbalanced articles” (Bero et al. 1992; Rochon 1994) (p. 159).
• Symposia, 1965–93:
• 11 published symposia on passive smoking (6 journal special issues, 5 books). None was peer reviewed.
• Six were sponsored by the industry or affiliates (CIAR, the Tobacco Institute, Fabriques de Tabac Réunies). Two of those six did not acknowledge the sponsorship, and some sponsorship was routed through independent organisations.
• They were held on four continents. CTR paid for talks and travel.
• Bero et al. (1994) comparison (p. 159):

The text says symposium articles were “significantly more likely” to agree with the industry position (and so on). Symposium authors compared with consensus-report authors: proportion of their publications that were peer reviewed, 71% vs 81%; industry-affiliated 50% vs 0%. - How the symposia were used: cited “as if they are peer-reviewed articles and balanced reviews”. Used against risk assessments and in PR, and presented as the view of “leading experts from around the world” (p. 159). - Quality (Barnes & Bero 1997): the chapter first concedes that “the issue of peer review and study quality is a contentious subject” (p. 160). Original research articles on SHS, 1980–94, were compared. Finding: “peer-reviewed articles were better quality than symposium articles independent of their source of funding, their conclusions… and the type of study design” (p. 160). - Reviews (Barnes & Bero 1998): - The a priori hypotheses were that “not harmful” reviews would be poor in quality, published in non-peer-reviewed symposia, and written by industry-affiliated authors (p. 160). Only the third was supported. - In 106 reviews, “the only factor associated with concluding that passive smoking is not harmful was whether the author… was affiliated with the tobacco industry”. - Table 7.1 (p. 161), odds ratios (95% CI): - quality score 1.5 (<0.1–67.5), p = 0.83; - not peer reviewed 1.3 (0.3–5.4), p = 0.70; - tobacco industry sponsored 88.4 (16.4–476.5), p < 0.001; - lung cancer vs other outcomes 1.6 (0.2–10.3); - heart disease vs other outcomes 1.6 (0.2–14.7); - year of publication 1.1 (0.9–1.3). - The text glosses this as “about 90 times more likely to be funded by the tobacco industry” (p. 160) and concludes that sponsorship “appears to influence the conclusions… independent of methodological quality”. - Note the chapter uses three terms for the same variable: the text says author “affiliated with” the industry, the table row says “Tobacco industry sponsored”, and the gloss says “funded”. - The industry’s publication-bias argument (pp. 160–161): - The industry claimed publication bias invalidates independent reviews. - Bero cites a preliminary study (cited as “Bero et al. (2004)”; probably the 1994 reference) showing about 20% of published articles are non-significant. - Misakian & Bero (1998): non-significant studies “take about two years longer to be published”; for human studies only significance predicted delay. - “Thus, the tobacco industry’s argument that statistically non-significant results are not published is invalid”. Recommendation: include unpublished data and update regularly, as Cochrane does.

7.3.4 Strategy 4: suppress (p. 161)#

• Lawyers and editors edited or blocked externally funded publications.
• The code word “zephyr” was used for “cancer” (BAT 1956).
• INBIFO’s sidestream-toxicity findings went unpublished.
• Additives: internal research on additives to “reduce, mask or otherwise alter the visibility, odour, irritation or emission” of SHS. Some increased carbon monoxide, N′-nitrosonornicotine and benzo(a)pyrene (Connolly et al. 2000). “Virtually none of this research has been published”, and the data are “not typically available to public health policymakers”.

7.3.5 Strategy 5: criticise (pp. 161–162)#

• “Science is improved by constructive criticism. However, the tobacco industry has misused legitimate means of scientific debate” (letters, editorials).
• Less legitimate methods: “attacking the integrity of researchers or obtaining data through lawsuits and reanalysing it using inappropriate techniques”.
• Letters were cited as if peer reviewed, and affiliated writers “often fail to disclose”. This supports calls for disclosure and peer review of letters (Rennie 1993).
• Paid consultants worked through conferences, publications, media, trial testimony, “forming a scientific society on indoor air”, and statements to government committees, internationally.

7.3.6 Strategy 6: change scientific standards (pp. 162–163)#

• Targets: “sound science” (Ong & Glantz), risk-assessment standards (Hirschhorn & Bialous), international product standards (Bialous & Yach), and data laws (Baba et al.).
• ISO/CORESTA: the industry “suggested the standards that were adopted” for tar and nicotine yields. These “incorrectly imply that there are health benefits from low-tar and low-nicotine products” (p. 162).
• ASHRAE: the industry had been involved in ventilation standards “for over 20 years”. By “generating data and presenting it to the committee” it influenced ASHRAE, which “resulted in a standard that ignores the health effects of second-hand smoke exposure, concentrating instead on a ‘comfort’ standard” (p. 162).
• “Sound science” and GEP (early 1990s):
• “All scientists agree that research should be rigorously conducted. But the ‘sound science’ and ‘GEP’ campaigns were public relations efforts controlled by industry executives and lawyers to promote unreasonably high standards of proof”.
• Example: the claim that epidemiology “can never establish evidence of harm because [it] cannot ‘prove’ causality”. This ignores whole-evidence assessment (p. 162).
• Low-dose risk assessment: the industry “also developed a campaign to criticise the technique of risk assessment of low doses of a variety of toxins” (Hirschhorn & Bialous 2001) (p. 162).
• Coalition: the industry worked with the chemical, petroleum, plastics and chlorine industries to develop these criticisms. “The first version of GEP was drafted by the Chemical Manufacturers Association.” By the late 1990s the PR campaign ended and the strategy moved to legislation.
• Obtaining others’ data: Philip Morris’s steps were to ask, then litigate, then push data-release policies (p. 162).
• US legislation (p. 163):
• A 1998 data-access rider (FY1999 omnibus) made federally funded research data available through FOIA “for the first time”.
• The Data Quality Act (2000; FY2001 rider) required OMB and agency guidelines to “ensure and maximise the quality, objectivity, utility and integrity of information”. “The standards to be adopted were created by the industry sponsors, not independent researchers.”
• Both were “initially passed and adopted without a legislative hearing, committee review or debate”, although the public could comment on implementation (Renner 2002). AAAS and others raised concerns about confidentiality, deterring research subjects, misinterpretation, delay, security, and administrative burden.
• Documents show the laws were “driven by tobacco industry efforts to coordinate corporate interests”, with the industry’s involvement hidden.
• The 14-point campaign plan (Baba et al.):
  • a sponsored poll on data access and epidemiological rules;
  • leveraging allies and groups already committed on the issue;
  • using scientists and technical conferences;
  • a small group of members of Congress;
  • the Administration;
  • allied industries (“fishing, utilities, waterworks”);
  • coalitions on other epidemiological issues (fishing industry, mercury, methylene chloride);
  • “sound science v. junk science” business education;
  • state laws;
  • “bipartisan support for ‘freedom of information’”;
  • lobbyists;
  • media;
  • business coalitions;
  • the Congressional Science Committee.
• Assessed consequences, mostly framed as “potential”:
  • challenging science “outside scientific journals”;
  • “potential to block agencies from using emerging science from non-industry sources and to slow the regulatory process”;
  • preventing or repealing policies;
  • shifting standards towards industry’s;
  • inequity: the laws “only apply to data generated with government funding, not industry funding”.

Panel 7.1 (Smith, Gilmore, Fooks): shaping risk assessment in the US and the EU (pp. 164–165)#

• US:
• The EPA’s Group A classification triggered Philip Morris’s “sound science” push for laws requiring epidemiological criteria before policy use, and FOI access to publicly funded data.
• It did not overturn the classification, but placed “a cloud over its validity” until 2002, “leading to delays in subsequent introduction of protective legislation”. There was “some success” with data-access laws and the Data Quality Act.
• Europe (mid-1990s):
• Philip Morris believed a similar campaign “might be even more effective in Europe, where officials had not yet taken up the scientific threat of ETS to the same extent”, and lobbied for mandatory GEP criteria (a term taken from the Chemical Manufacturers Association). These included disregarding relative risks below 2.0 “as too weak to warrant policy intervention”.
• As of late 2000, “no European Union resolution on GEP had been produced”. “As far as we are aware, this remains the case” (p. 164). So the RR < 2.0 rule was not adopted in the EU, as far as the panel knows.
• BAT (from 1995):
• BAT managers “studied the Philip Morris campaigns carefully and from 1995 onwards considered lobbying for” a mandatory requirement for “structured risk assessment”, believing it could prevent public smoking restrictions. “By this stage, the industry was well aware of the negative health impacts of second-hand smoke”.
• BAT 1995: “a legislated demand for structured risk assessment”, with strict evidence rules, would “remove the possibility of introducing public smoking restrictions that are based on risk claims”, and make it “apparent that ETS has not been proven to be a cause of disease”.
• Gap: the exact meaning of “structured risk assessment” is unestablished because the relevant documents “have been redacted” by the company (footnote 3).
• Discretion: “The challenge will be to persuade government departments to subordinate policy or judgemental considerations in favour of scientific rigour in risk assessment” (Gretton c. 1995). The panel’s gloss: “In practice, this constituted a way of undermining the precautionary principle” (p. 165).
• “Key innovation”: risk assessment inside “cost-benefit analysis”, used interchangeably with business impact assessment. This embedded economics and required protections “to be justified on the basis of economic costs” (p. 165).
• Route:
• Covington & Burling (1996) advised that, although there was “little interest in risk assessment within the European Commission at the time”, structured risk assessment might be inserted into detailed guidance for business impact assessment, flagged as a Commission priority in 1996.
• A campaign with “known links to the tobacco industry was unlikely to succeed”, so BAT was advised to use a “front group” and recruit other regulated firms.
• It approached the European Policy Centre (“with strong links to the Commission”). Together they recruited oil, chemical and pharmaceutical firms into an invitation-only “Risk Forum”.
• Claimed effect:
• The efforts “contributed to certain amendments to the Treaty on European Union (EU, 1997)”: a duty to “consult widely” and minimise the “burden” on “economic operators”. BAT interpreted this to mean that business impact assessment and risk assessment “were now mandatory within EU policymaking”, and called it “an important victory”. (The panel reports BAT’s own reading; it does not show that the Treaty text itself mandates risk assessment.)
• Commission impact-assessment guidance now includes risk assessment (EC 2009).
• Reinvention:
• In 2006–07, “under pressure to open up to civil society organisations and other members” of the European Policy Centre (then under new leadership), the Risk Forum companies left and set up the European Risk Forum. It calls itself “an expert led, not-for-profit think tank” “despite solely representing corporate interests, virtually all of which are connected to the chemical and tobacco industries”, which its chair confirmed in 2010.
• It urges “a more structured approach to risk assessment and risk management”, with unclear content.
• Hedged verdict: corporate efforts have been “somewhat successful” in redefining responses to risk, “including those that limit use of the precautionary principle in the EU (Löfstedt, 2004) and the United Kingdom (Dodds, 2006)”. But “it is not yet possible to assess the success” of the campaigns.
• Close: corporate input is “of course legitimate” if transparent, “not excessive in comparison to other stakeholders”, and not compromising public welfare (p. 165).

7.3.7 Strategy 7: lay press (p. 166)#

• Kennedy & Bero (1999): 180 newspaper and 95 magazine articles (North America, 1981–95).
• 66% of newspaper and 55% of magazine articles “left readers with the impression of continuing controversy”. The proportion “remained relatively constant” over the period.
• Industry-sponsored studies were rarely cited, but industry officials were. Of the newspaper articles, 52% cited industry officials, 56% government officials and 46% independent scientists. This “could have contributed to the emphasis on controversy” (hedged).

7.3.8 Strategy 8: direct to policymakers (pp. 166–167)#

• Case studies: two risk assessments, two state indoor-air rules and two federal tobacco rules (archives, dockets, transcripts, interviews).
• Notice-and-comment participation matters for legitimacy and “could help prevent the ‘capture’” of risk assessment (Wilson 1989).
• US EPA: development was “considerably delayed”. 64% (69/107) of submissions called the conclusions invalid; 71% (49/69) of those came from industry-affiliated individuals. They relied selectively on non-peer-reviewed literature, especially symposia.
• California: participation “was not balanced… and was dominated by the tobacco industry”. “Critics and supporters… used different criteria to evaluate the science, suggesting that they were constructing the evidence to support their predefined positions” (p. 166). This is a symmetric observation.
• Workplace rules (p. 167):
• The Washington and Maryland rules passed. The federal OSHA rule “failed”; the industry’s strategy there included “produce data to counter the findings”.
• State opposition from the industry, small businesses and business organisations “appeared to be coordinated”, often with undisclosed industry support.
• Non-science arguments dominated overall, but opponents used science arguments more than supporters and cited symposia or low-quality articles.
• Mechanism: “Apparent disagreement among experts during public testimony reinforces uncertainty”. Industry experts used “different criteria… different bodies of evidence” and argued “about specific studies rather than… the body of evidence as a whole” (p. 167).

Box 7.2: WHO Committee of Experts (CETID 2000), p. 168#

• Subversion of WHO was “elaborate, well financed, sophisticated and usually invisible”. WHO was “one of their foremost enemies”.
• The campaign “rarely” addressed the merits. It aimed to:
• divert attention;
• cut WHO budgets;
• “pit other UN agencies against the WHO”;
• persuade developing countries that tobacco control was a “‘first world’ agenda”;
• “distort the results of important scientific studies”;
• discredit WHO.
• It worked through front organisations, experts with hidden ties, and influence on other agencies and developing-country delegates.
• The Committee “believes” it “slowed and undermined” tobacco control. It is “reasonable to believe” it caused “significant harm”, a judgement it bases on the volume of subversion found “in its limited search”. The lives lost “may never be quantified”. It calls the executives’ conduct “unacceptable” and says it “must be condemned”.

7.4 Lessons learned (pp. 167–170)#

• Summary: consistent, international strategies; dissemination “despite the questionable conduct of much of this research”; lawyers in control; a need for “global awareness” (p. 167).
• Five diagnostic questions (p. 167): 1. Does the controversy exist only because interest-group-funded findings contradict others’ data? 2. Is it supported mainly by interest-group-supported publications? 3. Is it supported mainly by low-quality research? 4. Is it perpetuated in the press by affiliated individuals? 5. Are the data suggesting controversy presented to policymakers only by the interest group?

Policymakers “should apply these questions to all situations in which a company has an interest in creating controversy”. - Stated limit: tobacco “differs substantially from other industries in the deadly nature of its products when used as directed, and the historical lack of regulation”, but its methods “are similar to those of other corporate interests” (p. 167). - Evidence for other sectors (pp. 167–169): - pharmaceutical sponsorship and favourable outcomes, with reasons “not clear” and not explained by quality; - the food industry (Nestle; Levine et al., Olestra); - the Union of Concerned Scientists (2007) on ExxonMobil and climate (doubt, front groups, sponsored spokespersons, “sound science”); - Markowitz & Rosner on the chemical, asbestos and lead industries (tetraethyl lead, asbestos, PVC). Note: the cited 1991 book is, per the reference list, on silicosis (“Deadly Dust”), so the “asbestos” attribution is not clearly supported by the works cited; - the 2005 IJOEH special issue on chemical companies using the same eight strategies. The chapter notes it “relies heavily on internal company documents that the authors obtained by serving as expert witnesses in litigation” (p. 169), so this evidence too comes through litigation, from authors who were party-side experts. - Caution: “The public health community must learn more about the internal behaviour of corporations other than the tobacco industry in order to make conclusions about similarities” (p. 168). - Disclosure (p. 169): - Sponsors’ roles “can be evaluated only if interest group involvement in all steps… is fully described”. But funding disclosure “often provides incomplete information”. - “If internal tobacco industry documents had not been made available… much of what is known… would have remained undiscovered.” - Recommended scope: - sponsors’ roles at every stage; - personal financial ties (consulting, stock, honoraria), which are increasing and linked to favourable outcomes even without sponsorship; - ties of critics of harm research; - disclosure in articles, letters and editorials, press citations, and policy or legal testimony. - Peer review should be expert and sponsor-independent. Risk-assessment panels should weigh “affiliation or interest group bias” alongside competence and diversity (von Winterfeldt). Transparency “will not prevent their involvement” but helps to reveal strategies. - Limits of disclosure (pp. 169–170): - “necessary, but not sufficient”. It may deter conflicted ties and build trust (Cho 1998). - Krimsky: disclosure is a “rationalisation for creating more serious conflicts” and a “‘public relations’ response”. - Transparency “will not eliminate the biases”. “Eliminating financial ties may be the best way” if industry research is seen as less credible. - Options (p. 170): - a total ban on “clinical investigators’ financial ties to companies that fund their research” (Krimsky; Dana), which would remove the need for oversight committees to “manage” conflicts; - “sequestration”: an independent institute funded by companies (Schafer); - the objection that elimination “may not be realistic today” (Shamoo & Resnik); - self-regulation (Boyd et al.). - Tobacco: - Existing bans at Harvard (School of Public Health, per the reference) and the University of Sydney; the Legacy Foundation requires them. - “Bans on tobacco industry support for research are warranted in view of the industry’s history of deception”, further justified by motives: “distracting attention…, gaining credibility and using the research for public relations” (p. 170).

Table 7.2: key dates (EEA-compiled; Keating 2009; Ong & Glantz 2000), p. 171#

The table covers active smoking almost entirely: - 1604: King James I. - 1908: UK under-16 sales restriction. - 1931: Roffo, tar tumours in rabbits. - 1936: Ochsner. - 1938: Pearl. - 1939: Müller. - 1930–41: Schairer & Schöniger. - 1942–44: Jena dissertations. - 1946: Stocks, “startling” six-fold rise. - 1947: MRC meeting discusses road tar, urban air pollution, traffic fumes and smoking. The table’s phrase “factors which prepare the soil rather than sow the seed” is cited to Tudor Edwards (1946) and Keating, so it is not clearly the MRC’s own wording. - 1948: Doll & Hill preliminary; the inhalation anomaly fed Fisher’s denial. - 1950: five papers “largely doubted and generally ignored”. - 1953: UK advisory committee: “association was causal”, “young people should be warned”. - 1954: Doll & Hill doctors’ study; Hammond & Horn; the industry’s “1954 emergency” and Hill & Knowlton’s campaign to “‘manufacture doubt’… without actually denying it”. - 1964: US Surgeon General. - 1970s–80s: first SHS studies; 1986 causal. - 1993–98: “Tobacco industry subverts the WHO International Agency for Research on Cancer (IARC) study”. This is the chapter’s only mention of interference with the IARC study; the body (p. 153) reports that study’s results without mentioning it, and gives no detail or source beyond the table’s “Ong and Glantz, 2000” (not in the reference list).

Case timeline#

Chapter-sourced events (report pages in brackets)

[External] context for dating effective action and harms. The chapter supplies little of this. - Harm magnitude: the chapter gives no burden estimate. Öberg et al. (2011, Lancet; verified via Europe PMC) estimated 603,000 deaths worldwide attributable to SHS in 2004, about 1.0% of global mortality, 28% of them in children. This was published before the chapter and not cited in it. - IARC 1998 study (Boffetta et al., JNCI 90:1440–50; verified): spousal OR 1.16 (95% CI 0.93–1.44); workplace OR 1.17 (0.94–1.45); childhood exposure OR 0.78 (0.64–0.96). - Unverified, from general knowledge: - WHO FCTC adopted 2003, in force 2005. Article 5.3 protects policy from tobacco-industry interests; Article 8 covers protection from exposure. - IARC Monograph 83 (2004) classified involuntary smoking as a Group 1 carcinogen. - The US Surgeon General (2006) found “no risk-free level” of exposure. - US v. Philip Morris (2006) RICO judgment. Point-of-sale corrective statements ran from October 2023 to June 2025 (verified via Europe PMC abstracts). - First national comprehensive smoke-free workplace laws in Europe: Ireland, 2004; England, 2007. - EU Council Recommendation on smoke-free environments, 2009. - US OSHA withdrew its indoor-air proposal in 2001.

Lag (my inference from the above; not computed in the chapter): - First major published SHS warning (1981) to authoritative causal judgement (1986): about 5 years. - First warning to first comprehensive national smoke-free workplace laws in Europe (from 2004): about 23 years. - A US federal workplace rule was never adopted. - Industry knowledge ran ahead of public admission. There was internal INBIFO work on sidestream toxicity (INBIFO was bought in 1970; the chapter does not date the specific studies), and the panel says BAT “was well aware” of SHS harms by 1995 (p. 164). - Delays the chapter attributes to industry tactics (none is measured in time): the EPA release (“probably delayed”, p. 153; “considerably delayed”, p. 166), the “cloud” over the EPA assessment that persisted “until 2002” (p. 164), and ten years to recover the suppressed Japanese data (p. 158).

The authors’ own lessons and conclusions#

Lessons derived from the evidence (analytical): 1. The industry’s research-manipulation strategies were long-standing, consistent across active and passive smoking, and international (pp. 152, 154, 167). 2. Lawyers and executives, not scientists, controlled research design, conduct and dissemination, partly to shield it from discovery (pp. 152, 155–156, 167). 3. Sponsorship is associated with conclusions independently of measured methodological quality (pp. 154, 160, 167–168). The chapter says the reasons “are not clear” (p. 167) and proposes that biased outcomes “may be the results of” question framing, conduct and (non-)publication (p. 168). The mechanism is a hypothesis; the association is the finding. 4. Sponsored, non-peer-reviewed publications (symposia, letters) were used as if they were peer-reviewed and consensus sources (pp. 159, 161–162, 166). 5. Controversy was sustained in the media partly by quoting industry officials as experts (p. 166). 6. In regulatory processes, the industry dominated comment periods and used different evaluative criteria and study-by-study critique to create apparent expert disagreement (pp. 166–167). 7. The industry shaped standards (ISO, ASHRAE), standards of proof (“sound science”, GEP) and procedural law (data access and quality) (pp. 162–163). Panel: it also shaped EU risk-assessment and impact-assessment architecture (pp. 164–165). 8. Other industries use similar methods, but more evidence on their internal behaviour is needed before firm conclusions (pp. 167–169). 9. Without litigation-forced document disclosure, most of this would be unknown (pp. 151, 169).

Recommendations (advocacy): - Users of risk data should interrogate the sources of controversy using the five questions (p. 167). Policymakers should apply these “to all situations” where a company has an interest in creating controversy (p. 167). - Full disclosure of funding, of sponsor roles at every research stage, and of investigators’ personal financial ties, in publications, letters, press citations and testimony (p. 169). - Critics of research on harms should disclose their ties (p. 169). - Peer review should be expert and independent of sponsors (p. 169). - Risk-assessment panels should consider affiliation and interest-group bias in selecting experts (p. 169). - Reviews should include unpublished data and be regularly updated, as Cochrane reviews are (p. 161). - Journals should require disclosure from letter writers and peer review letters (p. 162, via Rennie). - “Bans on tobacco industry support for research are warranted” (p. 170). More general bans or sequestration are presented as debated options, not endorsed (p. 170). - Global awareness of the industry’s strategies is needed, because its efforts are international (pp. 152, 167). - Panel 7.1: corporate influence on risk assessment must be transparent, not excessive relative to other stakeholders, and must not compromise public welfare (p. 165).

The chapter does not recommend specific regulatory measures on SHS itself (smoke-free laws). Its recommendations concern the integrity of the evidence and advisory process.

Mechanisms and dynamics#

1. Controversy as the product; delay as the goal. The engine is the mismatch between policy’s need for consensus and science’s normal uncertainty (p. 153). The internal documents show two aims: producing and publicising counter-evidence that passive smoking “is not harmful” (the Roper “antidote”, p. 154), and keeping the argument open: “keep the controversy alive” (p. 154). Controversy “is likely to slow or prevent regulation” (p. 154). Even failed challenges produced delay, per the chapter: the EPA release (“probably”, p. 153; “considerably”, p. 166) and the “cloud” over its validity that lasted “until 2002” (p. 164). The dynamic is time-asymmetric (my inference): the chapter says restrictions threatened sales (p. 153), so each year of delay protected revenue, while exposed non-smokers bore the health cost.

2. Controlling the research pipeline end to end. The chapter maps influence onto every stage of research: - Agenda: funding choices favoured confounders, alternative causes and other indoor pollutants, and exposure rather than effects (pp. 155, 157). - Design and conduct: run by lawyers (pp. 155–156, 158). - Analysis: selective exclusion of data (Yano, p. 158). - Publication: sponsored symposia (p. 159) and suppression (p. 161). - Synthesis: industry-affiliated reviews (pp. 160–161). - Translation: to the press and policymakers (pp. 166–167).

The key analytical claim, which the chapter hedges (“may be”, p. 168; reasons “not clear”, p. 167), is that bias enters through framing, conduct and publication, not detectably poor methods (pp. 154, 168). If so, standard quality appraisal will miss it; Table 7.1 (quality score not associated with conclusions) is consistent with this.

3. Legal architecture as an epistemic shield. Routing science through counsel created privilege. Lawyers screened reports, chose projects and controlled dissemination. The Wells memo shows the logic that knowledge is liability: “cease producing information… helpful to plaintiffs” (p. 156). The same legal system later became the mechanism of exposure through discovery and the MSA (pp. 155, 169). Litigation was both a motive for concealment and the route to disclosure. Visibility therefore depends on where litigation happens to occur (pp. 152, 168–169).

4. Borrowed and laundered credibility. The industry understood credibility as the scarce resource (p. 154; the Philip Morris funding criterion, p. 155). It acquired credibility in several ways: - boards of well-respected scientists, even though many projects bypassed them (p. 156); - a peer-reviewed stream on non-tobacco topics (p. 157); - a university professor with secret ties (p. 158); - named independent investigators on a lawyer-managed study (p. 158); - symposia that looked like journals (p. 159); - think tanks and “ostensibly independent” organisations (pp. 165, 168); - small-business coalitions with undisclosed support (p. 167); - self-descriptions such as “an expert led, not-for-profit think tank” (p. 165).

A disclosure that is technically accurate but uninformative (“financial support from several companies”, p. 158) is itself a laundering device.

5. Distraction and alternative causation. Research on genetics, stress, diet (p. 155), green tea (p. 158) and other indoor pollutants (p. 157) supplied alternative explanations and diluted attention. Exposure studies supported the claim that doses were too low to matter (p. 157).

6. Technical-fix substitution. Instead of restricting the hazardous activity, the industry promoted engineering controls and product tweaks: - ventilation: the fabricated study (p. 157) and ASHRAE’s “comfort” standard (p. 162); - additives to reduce the visibility and odour of smoke, some of which raised toxin emissions (p. 161); - machine-measured low-tar and low-nicotine yields implying health benefit (p. 162).

The pattern (my synthesis) is a fix aimed at perception and acceptability rather than harm, backed by sponsored data. Only the ventilation case is documented in the chapter as used against restriction: the fabricated study “was used to support testimony that indoor smoking restrictions are not necessary” (p. 157). For additives and low-tar yields, the chapter documents the research and the misleading implication, not their use as an alternative to restriction.

7. Raising and reshaping the standard of proof. This is the most direct link to the report’s precautionary theme. The industry moved from contesting studies to changing the rules for what counts as evidence: - “sound science” and GEP rhetoric that epidemiology cannot “prove” causation (p. 162); - a proposed threshold dismissing relative risks below 2.0 (p. 164). For an exposure with RR ≈ 1.2, as in the IARC study (p. 153), this rule would guarantee a finding of “not proven” (my inference; the chapter does not spell this out). The panel reports that no EU resolution on GEP was produced (p. 164); - mandatory “structured risk assessment” to make it “apparent that ETS has not been proven” (p. 164); - subordinating “judgemental considerations” to “rigour”, which removes regulators’ discretion to act on the weight of evidence (p. 164); - embedding cost–benefit or business-impact tests, which shifts the burden onto protective measures to justify their economic costs (p. 165).

The industry’s confidence that procedure would determine outcome is striking. BAT managers believed a legislated demand for structured risk assessment “would… remove the possibility” of restrictions (p. 164), and “wanted to use risk assessment as a way of limiting officials’ discretion” (p. 164). The documents, as the panel presents them, suggest a mental model in which whoever sets the evidentiary rules decides the result, and official discretion is an obstacle to be limited.

8. Procedural weapons and asymmetric transparency. - Participatory mechanisms designed to prevent capture were used to flood the process (EPA 64%/71%, p. 166; California dominated by the industry, p. 166). - Procedural litigation succeeded where scientific challenge failed (p. 153). - The data-access and data-quality laws used the language of openness (“freedom of information”) but applied only to publicly funded data, not industry data (p. 163). - Lawsuits were also used to obtain independent data for hostile reanalysis (pp. 161–162).

The chapter’s own remedy is transparency. The industry turned a selective transparency against its opponents. The crux is symmetry.

9. Resource asymmetry and coordination. Manipulation strategies are costly. Corporate groups can mount “expensive, coordinated efforts”; public-health groups “tend to act independently” and have fewer resources (p. 152). The industry coordinated across firms (the MSA defendants, CIAR’s founders), with other sectors (chemical, petroleum, plastics and chlorine on risk-assessment criticism, p. 162; oil, chemical and pharmaceutical firms in the Risk Forum, p. 165; fishing, utilities and waterworks listed in the data-law campaign plan as allies to mobilise, p. 163), and across jurisdictions (pp. 154, 157, 164). Tobacco often led while hiding behind broader coalitions (pp. 163, 165).

10. Portability, forum-shifting and reinvention. - The toolkit transferred from active to passive smoking (p. 154). - It moved from the US to Europe when officials there were judged less engaged (p. 164). - It moved from PR to legislation when the PR campaign ran its course (p. 162). - It moved to international bodies and developing countries, where WHO’s programme was framed as a “first world” agenda (p. 168). - Its vehicles were reinvented after exposure: CIAR became PMERP (p. 157), and the European Policy Centre’s Risk Forum became the European Risk Forum (p. 165).

11. Media dynamics. Coverage kept a steady share of “controversy” framing over 15 years (p. 166). Industry officials were quoted about as often as government officials (52% vs 56%) and more often than independent scientists (46%) (p. 166). The chapter does not explicitly invoke journalistic “balance” norms. Its data are consistent with the idea that quoting “both sides” kept the controversy alive even as expert consensus firmed.

12. How evidence is argued in policy fora. Industry experts used different criteria and different bodies of evidence, and argued “about specific studies rather than… the body of evidence as a whole” (p. 167). Critics and supporters alike constructed evidence around predefined positions (p. 166). The industry’s advantage was to make apparent disagreement stand in for real scientific disagreement (p. 167).

13. Industry mental models (from the documents). - Research is instrumental: PR, “legal responsibility”, “political need”, rapid rebuttal of “propaganda of the avid anti-smoking groups”, and recruiting witnesses (Pepples, p. 156). - Critics and WHO are enemies (p. 168). - Information is legal exposure (p. 156). - Credibility must be managed while manipulating (p. 154). - Procedural rules can foreclose policy outcomes (p. 164).

The panel states that by 1995 the industry “was well aware” of SHS harms while trying to influence the evidence base (p. 164). This was not a blind spot. The chapter portrays deliberate, informed strategy rather than self-deception.

14. Regulators’ and scientists’ positions. - The EPA revised its report in response to “valid criticisms”, won Science Advisory Board approval, and was “probably” delayed by the volume of comments (p. 153). - US notice-and-comment was open by design and so exploitable (p. 166). - State agencies succeeded despite opposition (p. 167). - Scientists appear as recruited consultants (pp. 154, 162), as a “sponsored agent” (p. 158), and as named investigators kept in the dark (Yano, p. 158). - Independent scientists produced the counter-evidence: meta-research on symposia, reviews and publication bias (pp. 159–161). That meta-research is itself a governance tool.

15. Distribution of costs and benefits. Harms fall on people who did not choose exposure: partners, children, and hospitality and other workers (p. 151). The benefits (sales) went to producers (p. 153). The WHO box adds a global-justice twist: the industry told developing countries that tobacco control served rich-country interests at their expense (p. 168). The chapter does not quantify harms or costs. Panel 7.1 notes that cost–benefit framing itself was promoted as a tool to raise the bar for protection (p. 165).

16. Framing and language. The documents’ vocabulary shows the framing strategies: - “sound science” vs “junk science” (pp. 162–163); - “good epidemiological practices” (p. 162); - “comfort” standard (p. 162); - “structured risk assessment” (p. 164); - “burden” on “economic operators” (p. 165); - “freedom of information” (p. 163); - “expert led, not-for-profit” (p. 165); - the code word “zephyr” (p. 161); - “a ‘first world’ agenda” (p. 168).

These are neutral or virtuous-sounding terms (rigour, openness, comfort, expertise, economic prudence) put to work for delay.

17. Innovation. The chapter says almost nothing about innovation in the report’s sense. What little appears concerns product modifications aimed at acceptability (additives, p. 161; low-tar, p. 162), not reduced harm. The word “innovation” appears only as the panel’s description of BAT’s lobbying tactic (p. 165).

Transferable insights (technology-neutral)#

1. When revenue depends on the absence of restriction, keeping a scientific question open can be a deliberate, resourced strategy whose goal is delay rather than refutation. Evidence: p. 153 (threat to sales), p. 154 (Roper, Boyse), p. 156 (Pepples), p. 164. Strength: strong for this case. The internal documents show intent, not just inference. As a general expectation for other sectors: moderate.

2. Sponsorship is associated with favourable conclusions even where measured methodological quality does not differ, so quality checklists alone may not detect it. The likely routes are question choice, conduct and selective publication. Evidence: pp. 154, 160–161 (Table 7.1: affiliation OR 88.4; quality and peer review not significant), pp. 167–168. Strength: strong for the association. Several datasets point the same way (though most are by the author’s group), and later Cochrane work agrees [External, verified: Lundh et al. 2017, RR 1.34 for favourable conclusions, “an industry bias that cannot be explained by standard ‘Risk of bias’ assessments”]. Suggestive for the mechanism: the chapter says the reasons “are not clear” (p. 167) and that bias “may be” the result of framing, conduct and publication (p. 168). Caveat: Table 7.1 measures affiliation (labelled “sponsored” in the table), broadly defined, and its CI is very wide.

3. Peer review does not filter out conclusions biased by conflict of interest. What predicts the conclusion is who wrote the paper, not where it appeared. Evidence: Table 7.1, p. 161 (peer review OR 1.3, not significant). This sits alongside the chapter’s emphasis on non-peer-reviewed symposia as vehicles (p. 159). Strength: moderate. It rests on one regression with small numbers, and the chapter itself emphasises the peer-review distinction more than this result warrants.

4. Controlling the research agenda (which questions are funded) can matter as much as controlling the answers. Funding alternative causes, exposure-only studies and “credibility” research on adjacent topics diverts and dilutes. Evidence: pp. 155, 156–157 (CIAR 70% non-tobacco; special-reviewed projects mostly exposure studies), p. 158 (green tea). Strength: moderate–strong. The funding patterns are documented, but the author hedges the motives (“possible”, “may have”).

5. Routing research through legal counsel can turn inconvenient knowledge into privileged or never-created knowledge. Compulsory disclosure (for example litigation discovery) may be the only way such patterns come to light. Evidence: pp. 155–156 (Wells memo; Brown & Williamson screening), pp. 152, 169; redactions noted in the panel (p. 164). Strength: strong for tobacco. It also implies an observability bias: sectors without forced disclosure will look cleaner than they are.

6. Credibility can be borrowed and laundered: eminent advisory boards, academics with hidden ties, independent-looking conferences, think tanks, allied associations, and vague funding acknowledgements. Evidence: pp. 156–159, 165, 167, 168. Strength: strong. There are multiple documented instances, including an institutional finding (the Geneva commission) and a Congressional finding of fabrication.

7. Beyond contesting individual studies, an interested party can try to change the rules of evidence and procedure. Examples are demands for “proof”, minimum effect-size thresholds (RR < 2), mandatory structured assessment, cost–benefit gating, and limits on regulators’ judgement. These function as anti-precautionary decision rules that default to inaction under uncertainty. (That such rules, once enacted in law, outlast any single dispute is my inference; the chapter does not argue durability.) Evidence: pp. 162–165. Strength: strong on intent (documents). On effect, mixed: in the US, data-access and data-quality laws were enacted and the EPA assessment was clouded until 2002 (pp. 163–164); in the EU, the GEP/RR < 2 rule was not adopted (p. 164), the Treaty change is only said to have been “contributed to”, and the panel says it is “not yet possible to assess” the campaigns’ success, citing only “somewhat successful” analyses (p. 165).

8. Transparency mechanisms can be designed or applied asymmetrically and turned into tools of delay. Openness rules that bind public research but not proprietary data tilt the field while sounding neutral. Evidence: pp. 162–163; p. 164. Strength: moderate. The documented intent and asymmetry are strong. The downstream effects are framed as “potential”.

9. Participatory and procedural safeguards favour the well-resourced. Comment periods meant to prevent capture can be flooded, and procedural litigation can succeed where scientific challenge fails. Evidence: p. 152 (resource asymmetry), p. 153 (court; EPA delay), p. 166 (64%/71%; California dominated), p. 167 (coordinated, undisclosed allies). Strength: strong for the cases studied. Generalisation is plausible but untested here.

10. Measurement and engineering standards are overlooked battlegrounds. Whoever writes the metric (yield measurement) or the design standard (comfort-based ventilation) can define “safe” in ways that sidestep harm, and can offer technical fixes as substitutes for restricting the hazard. Evidence: pp. 157, 161, 162. Strength: moderate. The chapter relies on secondary citations and gives few details. [External, unverified: later legal and treaty action against “light”/low-tar descriptors supports the ISO point.]

11. Conventional news practice can sustain the impression of controversy long after expert consensus forms, particularly when interested parties are quoted as experts. Evidence: p. 166. Strength: moderate. It rests on one content analysis, and the causal link is hedged (“could have contributed”).

12. In adversarial policy settings, attacking studies one by one and using shifting evaluative criteria creates apparent expert disagreement. Weighing the body of evidence as a whole counters it. Evidence: pp. 162, 166–167. Strength: moderate–strong. There are multiple case studies, though all by the author’s group.

13. Once developed, influence strategies are portable across issues, jurisdictions and institutions, and exposed vehicles get reinvented. Examples: from active to passive smoking; from the US to the EU; from PR to legislation; from CIAR to PMERP; from the Risk Forum to the European Risk Forum. Evidence: pp. 152, 154, 157, 162, 164–165, 168. Strength: strong within tobacco; moderate across industries. The author herself says more evidence is needed (p. 168).

14. Producers can know more than regulators and the public, and can choose what to reveal. Examples: unpublished sidestream toxicity, unpublished additive emissions, and awareness of harms while contesting them. Evidence: pp. 158, 161, 164. Strength: strong for tobacco.

15. Disclosure is necessary but not sufficient. It can even legitimise conflicts. Where a sponsor has a record of deception, structural separation (bans or sequestration) may be warranted, though its feasibility is contested. Evidence: pp. 169–170. Strength: asserted/moderate. The chapter lays out the debate, and the tobacco-ban recommendation follows from the documented deception. It offers no evidence on how well any remedy works.

16. When risk evidence looks contested, ask where the controversy comes from: who funds it, where it is published, its quality, who is quoted, and who brings it to decision-makers. Evidence: p. 167. Strength: suggestive. It is a plausible heuristic but not validated in the chapter. If misused, it could dismiss valid industry-funded evidence purely because of its source.

17. Valid criticism and a large volume of strategic criticism can arrive together. The EPA fixed “valid criticisms” and was still “probably” delayed by the “sheer volume” of comments (p. 153). Legitimate critique and strategic obstruction can arrive in the same envelope. Evidence: p. 153, p. 166. Strength: moderate. The chapter’s own nuance supports it, but it gives no measure of how much criticism was valid.

Limitations, contestation and bias check#

Advocacy vs analysis - The chapter’s empirical core is strong. It rests on peer-reviewed document analyses and meta-research, much of it the author’s own, and on internal industry documents quoted directly. - Its prescriptions (bans on tobacco funding, full disclosure) are advocacy. They are argued explicitly and are proportionate to the documented deception. - The opening EEA box generalises most boldly (“other industries are drawing inspiration…”, p. 151). The author is internally inconsistent on this point: she asserts that the industry’s efforts “are shared by other corporate interests” (p. 152) and that its methods “are similar to those of other corporate interests” (p. 167), then cautions that “The public health community must learn more…” before drawing conclusions about similarities (p. 168). - The cross-industry evidence is thin and partly second-hand. The climate example rests on a single advocacy-organisation report (Union of Concerned Scientists 2007), reported without qualification (p. 168). The Markowitz & Rosner citation for “asbestos” includes a 1991 book on silicosis (per the reference list) (p. 169).

Self-citation and independence - About 34 of the references include Bero, so the chapter is effectively a curated account of one research programme. - The findings are broadly corroborated by other groups (Glantz, Chapman, Gilmore, Markowitz & Rosner, the WHO CETID) and by later reviews. Readers should still note that alternative readings of particular episodes are not presented.

Selection and observability bias (acknowledged) - The author notes that industry examples predominate partly because of resources and partly because litigation made documents public (p. 152), and that this insight “is not available for most corporate sectors” (p. 169). - This cuts both ways. It may understate manipulation in sectors whose documents remain private. It also means the eight-strategy template has been checked mainly against sectors that were litigated (tobacco, asbestos, lead, chemicals).

Asymmetry of scrutiny - The chapter concedes that public-health groups also construct evidence to suit their positions (pp. 152, 166), but examines only industry conduct. - It does not consider whether any public-health claims overreached. [External, unverified: for example, the 2005 Cal-EPA breast-cancer finding (p. 153) was not adopted by the 2006 US Surgeon General, who judged the evidence “suggestive but not sufficient”. A fair treatment would note that.]

Presentation of the science - The IARC 1998 result is given as “a 16 % increase… and a 17 % increase” (p. 153) without saying that both confidence intervals include 1 [External, verified: 0.93–1.44 and 0.94–1.45], or that childhood exposure showed no increased risk [External, verified]. [External, unverified: the industry publicised this non-significance.] Table 7.2 (p. 171) says only that the industry “subverts” the IARC study. The chapter would be more robust had it explained why modest, individually non-significant relative risks can support causal inference in the context of pooled evidence and toxicology. - More generally, the chapter never engages the scientific substance of the RR < 2.0 argument or of the misclassification critique. It treats both mainly as strategy. In my judgement misclassification was a real methodological question; the chapter’s own account concedes that some industry criticisms of the EPA assessment were “valid” (p. 153), though it does not say which.

Overstatements and slippages - The OR of 88.4 is glossed as reviews being “about 90 times more likely to be funded by the tobacco industry” (p. 160). An odds ratio is not a probability ratio. The variable is also affiliation (in the original paper, having received industry funding or taken part in industry-sponsored activities, which could include presenting at an industry-sponsored symposium), not funding. The CI (16.4–476.5) signals small cells. [External, verified: 39 of 106 reviews concluded no harm, 29 of those by affiliated authors.] The direction of the finding is robust; the headline magnitude is imprecise. - “The tobacco industry’s argument that statistically non-significant results are not published is invalid” (p. 161) is fair against that literal claim, but the word “invalid” understates the problem. The same study found non-significant results were delayed by about two years [External, verified: median 5 vs 3 years; 14 of 61 studies unpublished at the time]. Time-lag bias is a real form of publication bias. The chapter itself acknowledges this in the next sentence, recommending that reviews include unpublished data and be updated. The industry’s strong claim was wrong, but the underlying concern was not baseless. - “The standards to be adopted were created by the industry sponsors, not independent researchers” (p. 163) is compressed, and sits awkwardly with the chapter’s own statement that the Act “requires the Office of Management and Budget to develop” the standards (p. 163). The documented claim (Baba et al.) is that the industry drove the legislation and sought to shape the standards. [External, unverified: the data-quality guidelines cover information agencies disseminate, which can include third-party data, so the “only… government funding” equity point (p. 163) fits the data-access rider better than the data-quality law.] - The main text reports the 1998 court loss (p. 153) without its later fate [External, unverified: vacated on appeal in 2002]; the panel alludes only to a “cloud” lasting until 2002 (p. 164). - The chapter is inconsistent about the EPA delay: “probably delayed” (p. 153) vs “considerably delayed” (p. 166). - INBIFO is described as purchased by Philip Morris (p. 158) and later as “the German tobacco industry-supported research organisation” (p. 161). - Sponsorship–outcome association is attributed partly to “Barnes and Bero, 1997” (p. 154), which is actually a study of symposium-article quality.

Panel 7.1 causal attribution - Attributing EU Treaty and impact-assessment changes to BAT’s campaign is hedged (“contributed to”, “somewhat successful”, “not yet possible to assess”), which is appropriate. The Better Regulation agenda had many drivers. - [External, reviewer’s reading of the cited source, not re-checked in this audit] Löfstedt (2004) describes the EU’s shift from precaution towards regulatory impact analysis. Citing him as evidence of industry success in limiting precaution arguably reads more into that source than it claims. - The panel does not argue that risk assessment or impact assessment is bad policy per se. It argues that a self-interested party promoted it to foreclose protection. Readers should keep “who promoted X and why” separate from “is X good governance” to avoid a genetic fallacy.

Omissions (the chapter is largely a lightly updated 2005 article) - No mention of the WHO FCTC (2003) or its Article 5.3, the main governance response built on exactly this evidence. - No mention of the 2006 US v. Philip Morris judgment. - No mention of the 2004 IARC Group 1 classification or the 2006 Surgeon General report. - No engagement with the Enstrom & Kabat (2003, BMJ) cohort study, a peer-reviewed study that reported no significant association [External, verified from the abstract: spousal-smoking relative risk for lung cancer 0.75 (0.42–1.35) in men and 0.99 (0.72–1.37) in women. External, unverified: the paper disclosed partial tobacco-industry funding and was heavily criticised on exposure misclassification]. It is a salient test case for the chapter’s argument about peer review and sponsorship. - No quantification of SHS harm or of economic costs. No account of when effective protection arrived. No treatment of the hospitality-economics arguments that dominated much of the policy fight. [External, verified: Scollo et al. 2003 found all studies concluding negative economic impact of smoke-free policies were tobacco-industry supported, which would have strengthened the chapter.]

Table 7.2 (EEA-compiled) and citation errors - The 1939 Müller row appears to repeat language from the 1950 Doll & Hill row (“16 times”, “‘one in a million chance’”). This is probably a transcription error. - “Keeting” is a typo for Keating. - The source “Ong and Glantz, 2000” is not in the reference list. - “Bero et al. (2004)” in the text should probably be 1994. - References to Michaels (2008), Oreskes & Conway (2010) and White & Bero (2010) appear in the list but are not cited in the text. - Minor text typos: “Bero and Glatz, 1993” (p. 160); “Conolly et al., 2000” in the text vs “Connolly” in the list (p. 161); “Diethhelm” (p. 158); “N’-nitrosoniornicotine” and “methyline chloride” (pp. 161, 163).

The table relies on a biography and should be treated as orientation, not as a primary chronology.

No dissenting or industry panel. Given the documented deception this is defensible, but it means contested episodes are told from one side only (for example, Lee’s position on the Japanese study).

Hindsight bias. This is less of a concern than in most chapters. The core evidence is contemporaneous internal intent, not a retrospective judgement that “they should have known”. The risk lies elsewhere: reading every 1990s methodological dispute as purely strategic, when some were live scientific questions.

Pro-precaution framing. The chapter itself hardly invokes precaution; only the panel does (pp. 164–165). Its argument is about the integrity of evidence and process, which supports precaution indirectly. If evidence and standards of proof can be manipulated, waiting for “proof” is not a neutral stance. This framing is well supported by the documents. Readers should still distinguish two claims: - the strong claim: the industry deliberately exploited high proof standards; - the weaker, contestable claim: any demand for structured, quantitative or cost-aware assessment is suspect.

Notable quotes#

1. “to keep the controversy alive” (Philip Morris proposal, industry meeting notes, 1988; quoted p. 154)
2. “Tobacco industry lawyers and executives, rather than scientists, have controlled the design, conduct and dissemination of this research.” (p. 152)
3. “Since data do not ‘speak for themselves’ interest groups can play a critical role in generating and communicating the research evidence on risk.” (p. 152)
4. “Policymaking is facilitated by consensus… Scientific research, on the other hand, is characterised by uncertainty.” (p. 153)
5. “Originally, CTR was organized as a public relations effort.” (Pepples memo, 1978; quoted p. 156)
6. “cease producing information… that is helpful to plaintiffs” (Wells memo, 1986; quoted p. 156)
7. “…public relations efforts controlled by industry executives and lawyers to promote unreasonably high standards of proof…” (on “sound science”/GEP, p. 162)
8. “remove the possibility of introducing public smoking restrictions that are based on risk claims” (BAT, 1995; quoted in Panel 7.1, p. 164)
9. “to subordinate policy or judgemental considerations in favour of scientific rigour in risk assessment” (Gretton, c. 1995; quoted in Panel 7.1, p. 164)
10. “Detailed and accurate financial disclosures… are necessary, but not sufficient, for safeguarding the integrity of the research record.” (p. 169)

Open questions#

• Real-time discrimination: how can decision-makers tell legitimate methodological criticism from manufactured doubt before internal documents surface? The five questions (p. 167) depend on knowing sponsorship. How do they perform when sponsorship is hidden, which is the chapter’s own central finding?
• Generalisability: the author limits the analogy. Tobacco is uniquely “deadly… when used as directed” and was historically unregulated (p. 167). How far does the eight-strategy template apply to products or processes with substantial benefits, contested risks, and regulated markets? [External, verified: Legg, Hatchard & Gilmore (2021) synthesise evidence across eight sectors into a “Science for Profit” typology. Kearns et al. (2016) document an undisclosed 1960s sugar-industry-funded review. These suggest the pattern recurs, though the evidence is still document-dependent.]
• Remedies: what works? Disclosure, bans, sequestration and self-regulation are canvassed but not evaluated (pp. 169–170). [External, unverified: an industry-funded “independent” research foundation launched in 2017 is a live test of the sequestration idea and of CIAR/PMERP-style reinvention.]
• Symmetric transparency: can openness rules be designed to apply equally to public and proprietary evidence? The asymmetry flagged on p. 163 remains central. [External, verified: a 2021 US EPA “transparency in pivotal science” rule was vacated by a federal court and removed in June 2021. A May 2025 US executive order (“Restoring Gold Standard Science”, EO 14303) defines standards of reproducibility, transparency and uncertainty, requires availability of data underlying influential agency science while protecting confidential business information, and invokes the Information Quality Act. It places enforcement under a designated senior appointee. Whether this is a legitimate reproducibility agenda or a continuation of the “sound science” strategy traced here is a contested question for later analysis.]
• EU outcome of Panel 7.1: did impact assessment and “structured” risk assessment, with later developments, erode precaution in practice? [External, verified via Lauber & Brooks 2023, Globalization and Health: the European Risk Forum, renamed the European Regulation and Innovation Forum, lists among its self-declared achievements a role in the adoption of the EU “innovation principle” and the establishment of the Regulatory Scrutiny Board.] This is a direct continuation worth tracing from primary sources.
• Harm attributable to delay: the CETID box says the lives lost “may never be quantified” (p. 168). Could delay-attributable harm be estimated, for example from the timing of smoke-free laws against SHS burden estimates?
• Comment-process design: how should agencies weight mass or coordinated submissions, require disclosure of commenters’ funding, and prevent volume-induced delay?
• Legal privilege and health research: should privilege ever cover health and safety research relevant to public risk?
• Metrics and standards bodies: how are health-relevant technical standards (measurement, ventilation) insulated from producer capture?

Audit log#

Independent audit against the full text extract (PDF pp. 153–180) on 2026-09-25. I spot-checked the PDF itself for the bold emphasis in the Boyse quote and the Table 7.1 layout. Changes, one per line:

1. Authors: corrected the footnote. David Rosner is thanked for input on the manuscript, not as a collaborator on the cited studies. Listed the actual collaborators.
2. Authors: changed “later held” Cochrane roles to “has held” (still unverified).
3. Authors and Limitations: corrected the self-citation count from “roughly 30” to “about 34” (counted in the reference list).
4. Authors: added the “strategies outlined in this article” wording (p. 169) as further evidence of light updating from the 2005 article.
5. Opening box and Limitations: corrected the claim that the chapter body generalises to other industries more cautiously than the EEA box. The body also asserts it (pp. 152, 167) before calling for more evidence (p. 168). Flagged this as an internal inconsistency.
6. Authors: marked the Yano discussion-section claim as external and not re-verified.
7. 7.1: restored the hedge “likely to” on how interest groups communicate risk.
8. 7.2: replaced “The harm came from delay by volume” with the source’s hedge (“probably delayed”). Noted the unhedged “considerably delayed” on p. 166.
9. 7.2: marked the later appeal reversal of Flue-Cured v. EPA as external and unverified. The chapter does not report it.
10. 7.3: Boyse quote. Teams were proposed “in key countries” within each area. Gave fuller quote wording, identified the notes as a BAT document, and clarified what “emphasis added” covers.
11. 7.3.1: MSA “against” corrected to the source’s “between”.
12. 7.3.1: added Wells’s role, the bracketed “[British American Tobacco]”, and the chapter’s contrast between public “objective research” claims and lawyer control (p. 156).
13. 7.3.1: “About two-thirds” corrected to “almost two-thirds”. Clarified that the CIAR 70% and two-thirds figures are shares of projects, not of funding.
14. 7.3.3: added the pharmaceutical-symposia precedent (p. 159) and the chapter’s concession that peer review and quality are “a contentious subject” (p. 160).
15. 7.3.3: clarified the 71% vs 81% comparison (the share of authors’ publications that were peer reviewed). Noted “significantly more likely”.
16. 7.3.3: added that Barnes & Bero’s a priori hypotheses on quality and peer review were not supported. Noted the chapter’s inconsistent labels for the Table 7.1 variable (affiliated, sponsored, funded).
17. 7.3.6: ASHRAE sentence corrected. The “over 20 years” is the length of industry involvement, not a body of data.
18. 7.3.6: added the industry campaign against low-dose risk assessment (p. 162).
19. 7.3.6: added that the data laws were “initially” passed without hearings, though the public could comment on implementation.
20. 7.3.6: regrouped the 14-point campaign plan to match the source bullets (“allies and groups” and “scientists and technical conferences” had been mis-split). Added “fishing industry”.
21. Panel 7.1: Philip Morris’s Europe rationale now quoted exactly, not paraphrased as “less engaged”.
22. Panel 7.1: “BAT pursued” corrected to the source’s “considered lobbying for” structured risk assessment (also in the timeline).
23. Panel 7.1: added that the GEP/RR < 2.0 rule was not adopted in the EU (per the panel).
24. Panel 7.1: added Covington & Burling’s “little interest in risk assessment” caveat, and BAT’s reading that the Treaty made impact and risk assessment “mandatory”. Noted that this is BAT’s interpretation.
25. Panel 7.1: added the reason given for the 2006–07 split (“pressure to open up to civil society organisations”).
26. Box 7.2: added the Committee’s “limited search” caveat and its condemnation language.
27. 7.4: added that the 2005 IJOEH special issue relies on documents its authors obtained as expert witnesses in litigation.
28. 7.4: flagged that the Markowitz & Rosner “asbestos” citation includes a 1991 book on silicosis.
29. 7.4: narrowed the Krimsky/Dana ban to “clinical investigators’ financial ties to companies that fund their research”, matching the source.
30. Table 7.2: the “prepare the soil” phrase is cited to Tudor Edwards (1946), not the 1947 MRC meeting. Rephrased the IARC-subversion note.
31. Timeline: split the 1990 INBIFO row (the unpublished sidestream work is undated in the chapter). Moved the EPA comment statistics to an undated pre-1992 row. Changed Yano “reverses” to “changes the conclusion”. Added the late-2000 no-GEP-resolution entry.
32. Lag section: delays marked as attributed but unmeasured, with the chapter’s differing hedges.
33. Authors’ lessons item 3: mechanism (framing, conduct, publication) marked as the chapter’s hypothesis. The chapter says the reasons “are not clear”.
34. Recommendations: added the call for global awareness (pp. 152, 167).
35. Mechanism 1: corrected “not to win the argument but to keep it open”. The Roper memo aimed to produce counter-evidence. Marked the sales and cost asymmetry as my inference.
36. Mechanism 2: added the chapter’s hedge on how bias enters research.
37. Mechanism 6: removed an invented quotation (“reasonable alternative”). Noted that only the ventilation case is documented as being used against restriction.
38. Mechanism 7: marked the RR ≈ 1.2 point as inference and noted that GEP was not adopted. Softened “discretion is the enemy” to wording grounded in the panel’s “limiting officials’ discretion”.
39. Mechanism 9: attributed each allied sector to its specific page and role (planned mobilisation vs actual coalition).
40. Mechanism 14: removed the unsourced “good faith” characterisation of the EPA.
41. Insight 1: corrected the page references (Roper and Boyse are on p. 154).
42. Insight 2: split the rating. Strong for the association; suggestive for the mechanism.
43. Insight 7: removed the unsupported “more durable” claim (now marked as inference). Changed the effect rating from “moderate” to “mixed”, with US/EU specifics.
44. Insight 17: removed the unsupported “small” criticism framing. Restored the “probably” hedge.
45. Limitations: added the thin cross-industry evidence (a single advocacy report on climate; the mismatched asbestos citation).
46. Limitations: marked “exactly the result the industry exploited” and “misclassification was a real question” as external or reviewer judgement.
47. Limitations: softened the critique of “invalid” publication bias. The chapter itself acknowledges time-lag bias in its next sentence.
48. Limitations: noted that the chapter itself says OMB develops the data-quality standards. Added an external, unverified caveat on the scope of the equity point.
49. Limitations: added the internal inconsistencies (EPA delay wording; INBIFO’s description) and the minor text typos. Renamed the “Table 7.2 errors” heading to include citation errors.
50. Limitations: marked the Löfstedt reading as the reviewer’s external judgement.
51. Digest: corrected “70% of peer-reviewed funding” to “70% of peer-reviewed projects”. Attributed the fabrication finding to the Congressional subcommittee.
52. Digest: restored the “probably” hedge on the EPA delay. Added the failed EU GEP rule, “considered lobbying”, the Table 7.1 label, and the unsupported a priori hypotheses.
53. Digest: softened the lesson and insight on how sponsorship biases research, changed insight 6’s rating to mixed, and added internal-inconsistency and cross-industry caveats.