Late Lessons, Jensen Huang and AI
Markdown version

LL2-04 digest — Ch4 Too much to swallow: PCE contamination of mains water#

Late lessons from early warnings: science, precaution, innovation (EEA 2013), report pp. 76–91. Author: David Ozonoff (environmental health, Boston University). Panels: Christina Rudén (KTH; regulatory toxicology) and Joy Onasch (Toxics Use Reduction Institute).

Core story#

In the 1960s Johns-Manville, an asbestos-cement pipe maker, developed a plastic lining for mains in low-flow areas with colour and taste problems (p. 77). The lining was applied as a slurry in the solvent PCE. A first trial in 1966 gave water a “slight chemical taste and odour”. In early 1968 in Providence, air trapped in a new clear-lined pipe smelled like chloroform or dry-cleaning fluid, and the water kept a “very slight” odour under static conditions. Routine tests (pH, alkalinity, hardness) could not see organic contaminants. The company blamed incomplete “curing” and, as far as the record shows, never tested for residual PCE (p. 77). The pipe was installed from 1969 to 1979. Leaching was “accidentally discovered” in 1976, after more than 700 miles had been laid in New England. Some supplies still need continuous remediation (pp. 76–77).

Was the harm foreseeable? Ozonoff reconstructs the pre-1970 record to answer (pp. 79–82): - Hookworm therapy (from 1925). Dog deaths at therapeutic doses; individual susceptibility; rare serious or fatal reactions not tied to dose. - Occupational warnings, 1930s–50s. Chronic low exposure could be worse than acute; symptoms appeared within “accepted” limits; a 1953 clinical series (at 200–400 ppm, at or above Dow’s proposed ceiling) contradicted Dow’s 1952 view that serious organic injury was unlikely. - Dow’s own openly published 1960s research. Slow excretion and accumulation, delayed liver effects, unexplained neurological effects at the lowest doses, and a warning that PCE’s “non-toxic” image encouraged careless use. - A 1969 rat study. Of 27 chemicals tested in pairs for acute lethality, PCE was the one most often involved in more-than-additive effects (4 of 9 deviating pairs).

Public health managers already knew that solvents contaminate water, and a drinking-water standard said water “should contain no impurity which would cause offence to the sense of sight, taste, or smell” (dated 1968 in the chapter). So the judgement that experts would not have welcomed any PCE in their water “is not merely a statement based on hindsight” (pp. 82–83). A worst-case calculation (saturated water, 2 L a day) gives a steady body burden of about 1 g, “close to the dose used to treat hookworm” (p. 83).

Ozonoff finds no evidence that Johns-Manville concealed anything about PCE, unlike its conduct on asbestos. “Assuming that knowledge was not hidden”, indifference is the plausible explanation: the science “never figured in the water mains product design”, and “nobody made them care” (pp. 84, 86). After 1970 carcinogenicity testing began, every result was contested, and industry “consistently denied” PCE was a human carcinogen: “The problem is no longer invisibility and neglect, but intense scrutiny” (pp. 86–88).

Panels#

Authors’ lessons#

Derived from the evidence: - Foresight from existing information was possible (pp. 77, 83). - Disputes are normal in science, but non-scientific interests (ideology, money, market share, reputation) can distort them. Findings are contested when “someone with the means” cares about the outcome, and neglected when “nobody with means cares” (p. 78). - Risk accepted for a medicine does not transfer to a use with no matching benefit (pp. 80, 83). - Missing epidemiology “would not have been a reason to delay action” by the norms of the time (p. 84). - Specialist-only knowledge, plus ignorance among workers and physicians, kept the hazard off the agenda (p. 84). - “Uncertainty favours the side of inaction” when large firms have high stakes (p. 86). - The science “has not been hidden. It has been ineffective in guiding and catalysing action”, and the dispute is “irresolvable within science itself” (pp. 76, 88).

Recommendations: mechanisms to force the production, sharing and publication of exposure and effects information; normative and legal duty-of-care requirements for employers and manufacturers; criminal and civil liability as the main lever, with any mechanism (including moral pressure and voluntary standards) judged by the “acid test” of whether it makes a company care (pp. 86, 88).

Advocacy, uncited: industry is “artificially and purposefully creating doubt” and “buying extra time” (p. 88).

Mechanisms#

Transferable insights (strength)#

  1. Monitoring built for familiar failures misses novel ones (p. 77). Moderate.
  2. Test the implications of the explanation used to dismiss an anomaly (pp. 77, 83). Moderate.
  3. Hazard knowledge often exists but is not consulted by downstream integrators (pp. 82, 84, 88). Moderate–strong.
  4. Acceptable risk belongs to the use and its beneficiaries, not the substance (pp. 80, 83). Strong.
  5. Beneficial use and familiarity create false presumptions of safety (pp. 78, 81–82). Moderate–strong.
  6. Rare, delayed, cumulative and interactive effects need scale-appropriate evidence (pp. 79–82). Strong.
  7. Attention follows stakes and means, producing either neglect or contestation (pp. 78, 88). Suggestive–moderate.
  8. The same data yield different verdicts depending on assessment policy and institution (pp. 84–85). Strong for divergence; moderate for cause.
  9. Decision triggers must be explicit, because science alone cannot settle interpretation (pp. 76, 85, 88). Moderate.
  10. Uncertainty favours the better-resourced party (p. 86). Suggestive: asserted.
  11. Indifference is enough for harm; only levers that change what a firm cares about work (pp. 84, 86). Moderate.
  12. Viable substitutes spread slowly without regulation (p. 87). Moderate: thin evidence.

Caveats#