LL2-07 digest — Ch7 Tobacco industry manipulation of research#

EEA Report No 1/2013, report pp. 151–178 (PDF 153–180).

• Author: Lisa A. Bero, sole author.
• Panel 7.1 (pp. 164–165): Smith, Gilmore and Fooks. Complementary, not a dissent.
• Box 7.2: quotes the WHO Committee of Experts on Tobacco Industry Documents (2000).
• Table 7.2: EEA-compiled.
• No industry response panel.

Core argument#

The chapter is not a conventional “early warning” case. It maps how an industry manipulated the production, synthesis and use of evidence on second-hand smoke (SHS) to delay regulation. The main sources are internal documents released through US litigation, together with the author’s own meta-research.

• Aim: the industry’s “primary motivation… has been to generate controversy” (p. 152).
• Engine: “Policymaking is facilitated by consensus” while science “is characterised by uncertainty” (p. 153). Controversy “is likely to slow or prevent regulation” (p. 154).
• Intent in the documents: research was the “antidote” to the passive-smoking issue (1978), and teams would “keep the controversy alive” (1988) (p. 154).
• Constancy and control: the strategies were constant from the 1950s to the 1990s, and were controlled by lawyers and executives, “rather than scientists” (p. 152).

Key evidence#

• Eight strategies (Box 7.1, p. 155): fund, hide, publish, suppress, criticise, change standards, press, policymakers.
• Funding bodies:
• The Council for Tobacco Research was set up for PR; lawyers controlled its funding from 1966 (p. 156).
• 70% of the Center for Indoor Air Research’s “peer-reviewed” projects examined non-tobacco pollutants; almost two-thirds of its executive-reviewed “special” projects concerned SHS, mostly exposure. A Congressional subcommittee found that one ventilation study it funded had “altered and fabricated” data (pp. 156–157).
• It was reinvented in 2000 (p. 157).
• Hidden sponsorship (p. 158): sidestream-toxicity findings kept unpublished; a lab coordinator found by a University of Geneva commission to be a “sponsored agent”; a lawyer-managed study that, per an original investigator’s reanalysis, excluded inconvenient data, with the full data changing its conclusion ten years later.
• Reviews: among 106 reviews, industry affiliation (labelled “sponsored” in the table) was the only predictor of concluding SHS “not harmful” (OR 88.4, 95% CI 16.4–476.5). Quality and peer review did not predict, contrary to the authors’ own a priori hypotheses (Table 7.1, p. 161). None of 11 symposia was peer reviewed (p. 159).
• Standards (pp. 162–163):
• The industry shaped tar-yield (ISO) and ventilation “comfort” (ASHRAE) standards.
• “Sound science” campaigns promoted “unreasonably high standards of proof”.
• US data-access and data-quality laws, pushed by the industry while hiding its role, apply only to publicly funded data (per the chapter).
• Panel 7.1: a proposed rule to discount relative risks below 2.0 (no EU resolution adopted, p. 164); BAT considered lobbying for “structured risk assessment” to “remove the possibility” of smoking restrictions; a front-group “Risk Forum” that “contributed to” EU Treaty “burden” language. Success “not yet possible to assess” (p. 165).
• Press and policy: 66% of newspaper articles conveyed “continuing controversy” (p. 166). In the 1992 EPA risk assessment, 64% of submissions called it invalid, 71% of those industry-linked. The volume “probably” delayed release (p. 153; “considerably delayed”, p. 166), though some criticisms were “valid” (p. 153).

Authors’ lessons and recommendations#

• Lessons: sponsorship is associated with favourable conclusions without visibly poorer methods; the reasons “are not clear”, but bias “may be” introduced through question framing, conduct and non-publication (pp. 167–168). Other sectors’ methods are said to be similar (pp. 152, 167), yet the author also says the field “must learn more” before concluding so (p. 168).
• Recommendations:
• five questions to trace the source of a controversy (p. 167);
• full disclosure of sponsors’ roles and personal financial ties in papers, the press and testimony;
• independent peer review (p. 169);
• disclosure is “necessary, but not sufficient”;
• bans on tobacco-industry research funding “warranted” (pp. 169–170).

Mechanisms#

• Doubt as product, delay as goal.
• Control of the research pipeline end to end.
• Legal privilege shielding knowledge.
• Laundered credibility.
• Distraction through alternative causes.
• Technical fixes offered instead of restricting the hazard.
• Raising the proof standard and removing regulators’ discretion (anti-precautionary by design).
• Asymmetric transparency.
• Flooding participatory processes.
• Reinventing vehicles once exposed.

Transferable insights#

1. Where revenue depends on non-regulation, sustaining uncertainty can be deliberate and aimed at delay (pp. 153–156). Strong (internal documents).
2. Sponsorship is associated with conclusions without detectable method flaws (pp. 154, 160–161). Strong for the association, later corroborated; suggestive for the mechanism (reasons “not clear”, p. 167).
3. Peer-review status did not predict conclusions; affiliation did (p. 161). Moderate.
4. Controlling which questions get funded matters as much as controlling results (pp. 155–157). Moderate–strong.
5. Legal structures hide knowledge. Forced disclosure is the main window, which creates observability bias (pp. 152, 155–156, 169). Strong.
6. Beyond attacking studies, an interested party may try to change evidentiary and procedural rules (pp. 162–165). Strong on intent; mixed on effect (US data laws enacted; EU RR < 2 rule not adopted; EU success “not yet possible to assess”).
7. Transparency can be applied asymmetrically and used as a weapon (p. 163). Moderate.
8. Procedural safeguards favour the well-resourced (pp. 153, 166). Strong for these cases.
9. Measurement and design standards are quiet battlegrounds (pp. 157, 161–162). Moderate.
10. News practice can sustain apparent controversy after consensus forms (p. 166). Moderate.
11. Influence toolkits are portable and reinvented (pp. 154, 157, 164–165). Strong within tobacco.
12. Disclosure is insufficient; structural separation may be needed (pp. 169–170). Asserted/moderate.

Caveats#

• The chapter is largely a synthesis of the author’s own work, updated lightly from 2005. It omits the WHO tobacco treaty (2003), the 2004 IARC classification, and the 2006 US court judgment.
• It does not quantify harm or date effective action.
• It reports the IARC 16% and 17% increases without noting that both confidence intervals include 1.
• It overstates “90 times more likely to be funded” (an odds ratio of affiliation, very wide CI). Calling the industry’s publication-bias argument “invalid” is fair against its literal claim but understates the two-year time-lag bias the chapter itself reports.
• Internal inconsistencies: EPA delay “probably” (p. 153) vs “considerably” (p. 166); other industries’ tactics asserted as “similar” (pp. 152, 167) yet said to need more study (p. 168). Cross-industry evidence rests partly on a single advocacy report (climate) and a mis-matched citation (asbestos).
• Panel 7.1 does not show that risk assessment is bad policy in itself.
• Table 7.2 has transcription errors.
• It concedes both sides “construct” evidence but examines only industry.
• Hindsight bias is limited, because the core evidence is contemporaneous internal intent.