Houdini Risk Assessments Discussed

Human and Ecological Risk Assessment: An International Journal,
Volume
9, Issue 5 September 2003 , pages 1129 - 1143
Does the Emperor Have Any Clothes: Using Mechanistic Information or Doing Houdini Risk Assessments?
Author: Mirer, Franklin E.1

Abstract: Risk assessment research rarely quells controversy. Mega-mouse, and mega-rat, experiments contradicted a threshold for carcinogenesis, yet thresholds are still argued. High to low dose continuity of response from cigarette smoking to environmental tobacco smoke, and from occupational asbestos exposure to take-home asbestos, contradict thresholds in people. Nevertheless, mechanistic hypotheses allege "Houdini Risk Assessments", which make risks disappear or allow industries to escape from protecting workers. Despite concerns for animal-to-human extrapolations, priority occupational exposures with sufficient or substantial evidence of carcinogenicity in people not addressed by new exposure limits include silica, sulfuric acid mist, chromates, diesel particulate matter, particulate matter generally, metalworking fluids, welding fume, and formaldehyde. "Houdini Risk Assessments" are exercises in "anti-hypothesis generation": ignore selected tumor sites and types; ignore data from people (as with formaldehyde and diesel); choose the most resistant species in laboratory tests; select biochemical parameters in which the most resistant species resembles people; assume a mechanism that gives threshold or steep exposure response for carcinogenic effect; and reduce estimated people risk by the parameter ratio to the most resistant species. NORA research should focus on quantitative reconciliation of laboratory and epidemiology studies, and develop a counter "anti-hypothesis" generation research agenda for key exposure circumstances.
>>>>>>>>>>>>>>>>>>
BrooklynDodger(s) comment: The Dodger(s) picked up and use(s) the framework term "Houdini Risk Assessment" from this paper the Dodger(s) found. Many papers are published exploring the mechanisms of action of toxic chemicals. The contribution of this new knowledge to public health inteventions depends in part on whether the hazard has been identified in people, or is being extrapolated from laboratory studies in animal species.

Where "mechanistic" research arises from hazard identification in animal studies, and limited or no data from people, it is frequently advanced as a "morning after pill" to abort risk assessments. The absolute and binary conclusion that a mechanism is "not relevent to people" (peroxisome proliferation associated rodent liver tumors, kidney tumors in male rats, urinary bladder tumors associated with calculi) stops risk assessment at the hazard identification stage and ends the gestation of exposure response assessment.

A less pernicious use of mechanistic data might be estimating population variability in extrapolating to low dose risk. Were the human population widely variable in the biochemical parameter associated with potency, then a flatter population exposure response relationship would be expected, leading to higher unit risk estimates at lower doses where the human risk can't be directly observed. This line of argument would apply where hazard identification is based on higher dose human data.