The Great Risk Assessment Debate
Comments by A. A. Brooks
During the last two months (
Although some participants stressed different facets, there were some agreements as expressed by Roth-Nelson in the attached message he has kindly permitted me to use. Dave Burmaster and Mike Marchlik have also permitted me to attach their messages, which reflect upon what can be only regarded as a retrogression of the applied risk assessment science. These messages raise serious questions about the risk assessment discipline:
This situation raises the question: Why does this hiatus concerning probabilistic methods exist in risk assessment and does not exist to any serious degree in other sciences?
Other sciences have not been immune to such differences but have resolved them in the space of a few decades. In spite of the fact the foundation work necessary to establish the probabilistic method is from 100 to 200 years old, perhaps more time is needed. In which case the debate should go on (at a more leisurely pace).
Many of these past differences have been resolved within the cloistered halls of science but some have not. Seldom has the difference involved the public and the government regulators to the degree that risk assessment currently does. If the latter is the cause then it is even more important that risk assessment as the application of the science be separate from the risk management as its decision making adjunct.
Roth-Nelson asks where this leads in the risk policy debate? I believe it must continue along two lines: 1) the resolution of the "safety factor" approach and the "probabilistic" approach and, 2) the freedom of the science to determine its own validity in an unbiased forum.
Personally, I intend to take EPA's best advice and try to get the laws changed to create an unbiased forum. In the meantime, I will always challenge deliberately biased estimates as being non-science and encourage an informed public involvement in the risk management process.
========= Attachment 1 =======================
Subject: DEFAULT UNCERTAINTY OR SAFETY FACTORS
Date: Thu,
From: Roth-Nelson <roth_nelson@compuserve.com>
To: brooks50@comcast.net
I was struck by apparent convergence in the ideas of
Charlie Shore says, where we apply as many as four orders of magnitude of uncertainty, the derivation of a quantitative risk value is unwarranted. I would add: Where we combine the compounding uncertainties of toxicity factors with those of highly exaggerated default exposure parameters, the potential overall uncertainty would surely reach a nonsensical level. Yes, we may assert we are at least placing extreme bounds on the potential range of risk--are they at all meaningful if the low and high ends are a million times apart?
What is the scientific merit in packaging and promulgating a chronic oral reference dose of, let's say, 1 mg/kg/d that in an extreme probability may fall very close to a LOAEL of 10,000 mg/kg/d. Consider, too, that EPA's boilerplate tenfold uncertainty factors (UFs) freely intermix factors of genuine uncertainty (animal-to-human extrapolation) with factors of natural variability (protection of sensitive subpopulations). The resulting hybrid number is a highly artificial RISK MANAGEMENT default value used to represent the high-end potential health risk from toxicity. Risk assessors everywhere may offer that high-end value as if it were a meaningful artifact of science.
Clark Carrington, while agreeing with Daniel Byrd there is indeed some empirical basis for generating default "safety" factors to set extreme boundaries on risk, wants a RISK ASSESSMENT value that explicitly scales the potential health risk without doctoring it up using modifiers of hypothetical uncertainty and variability. Presumably, that means we must no longer disguise the observed toxicity by shrinking the NOAEL or LOAEL to a tiny fraction or by always assuming as a default the zero-threshold, linear extrapolation of the high-dose cancer potency slope factor, which we have publicized as our best "science." Instead, we must follow a more objective risk assessment with a separate risk management exercise that weighs estimated uncertainty and variability and sets substantially arbitrary boundaries on acceptably "safe" levels of toxicity and exposure. At least by segregating our risk assessment biases from our risk management biases we may limit our propensity to mislead risk managers and the public.
Robert Lee agrees that "safety" factors are risk management values based on some level of scientific information, but a separate uncertainty analysis allows us to present the full boundaries or distributions of risk to risk managers. He would like to see the undoctored data plus separate but associated uncertainties and variabilities, not just a bunch of numbers altered by safety factors. Mike Buchman wants to see the whole body of evidence to estimate the boundaries on the probabilities of obtaining various UFs of 10. But he finds it is OK to interject societal values, such as safety factors, if they are explicit and first identified as separate from observed toxicity values. Alfred Brooks reminds us that using the unbiased data from human or animal studies does not preclude a subsequent risk management decision to protect the vast majority of a potentially exposed population, depending on the available resources for control or mitigation.
With the rise of discrete uncertainty and variability analyses, can the demise of boilerplate RfDs/RfCs and automatic zero-threshold, linear slope factors be far behind? Maybe it is high time risk assessors are instead furnished with standardized, empirical, central-tendency and high-end NOAELs/LOAELs and dose-response curves. Separate safety factors could be approximated that relate not to default assumptions but to the actual weight of evidence for each toxicant unless the body of evidence is too inconclusive. The risk management options would rest on the weight of evidence and its implications for high-end or central-tendency safety factors.
Meanwhile, the science numbers would stand alone from the policy numbers, thereby unmasking the often interchangeable use of the two. It is not so much a concern that safety factors are substantially arbitrary; it is their representation as reliably science-based rather than predominately policy-based numbers that so distorts the present practice of regulatory risk assessment.
Again, where does this thinking lead in the risk policy debate?
Wayne Roth-Nelson
==== Attachment 2 =================================================
Subject: Some Thoughts on "A Brief Introduction"
Date: Sat,
From: "David E. Burmaster"
<deb@alceon.com>
To: brooks50@comcast.net
Mike Marchlik wrote:
>My name is Mike Marchlik and I just joined
this list this past week. I
>liked the tail end of the discussion on risk policy. Back in 1985, a
>group of us at Ebasco used explicit uncertainty
analysis in risk
>assessment before the EPA boilerplate methods came out. We used it with
>the Cal Energy Comm for permitting MSW
incinerators, on Superfund sites,
>and at
>assessment came out, however, we found it difficult to get clients to
>use this approach.
>
>These tools have existed for a very long time and have been applied,
>published, debated and endorsed. Yet there is still a reluctance to use
>this information to guide policy decisions. So we spend large sums of
>money cleaning up to background levels, when a small portion of those
>funds would save many, many more lives in prenatal care and other
>underfunded social programs. Just
my thoughts on this issue.
>
>Mike Marchlik
Dear Mike:
cc RiskAnalListServ
Welcome to the listserv.
I agree! In 1984, the
Anderson et al, 1984 Anderson, E., N. Browne, S. Duletsky, and T. Warn, 1984, Development of Statistical Distributions or Ranges of Standard Factors Used in Exposure Assessment, EPA-68-02-3510, Revised Draft Final Report prepared for the Exposure Assessment Group of the US Environmental Protection Agency by GCA Corporation, Chapel Hill, NC, September 1984
Soon thereafter, the Agency underwent some kind of conversion to become stubborn and unthinking adherents of the present silliness that all adults weigh 70 kg AND drink 2 liter/day AND breathe 20 cubic meter/day AND ....
Effective immediately, these patently false "default" assumptions trumped all measurements and knowledge -- to the point that bad assumptions completely replaced pretty good (but not perfect) knowledge.
Worse, these patently false "default" assumptions basically stopped the pursuit of new knowledge, stopped proposed research, stopped inquiry, and often stopped clear thinking. We have now raised a generation of risk assessors who have simply learned to parrot untruths rather than explore, differentiate, and understand
1. what we know now
2. what we do not know now -- but we could learn if
...
3. what we cannot know
The Agency's unbending adherence to exposure factors that everyone (including school children) knows are false has made the Agency into the butt of many jokes and has demeaned everyone of us in the profession of risk assessment.
Yet, even with professional journals and conferences bursting with new applications and new methods for probabilistic (and fuzzy) analyses, many (but not all) people in the Agency have circled the wagons and continue to deny the obvious even more vehemently. This is especially true in regional and state offices.
At this time, the Agency has dug itself into such a deep, deep hole that one wonders how it can ever extricate itself.
An adage in computer science holds that "The purpose of computation is insight."
I say, yes!, and let's restore science to risk assessment by doing probabilistic computations that provide insight into variability in Nature and into uncertainty in human knowledge. If we continue to give distorted risk assessments to our risk managers, then the risk managers will continue to make distorted decisions.
Just noodling on the sad state of our profession Dave
**********************************************************
David E. Burmaster, Ph.D.
Alceon Corporation (R)
deb@Alceon.com
http://www.Alceon.com
(R) Alceon is a registered trademark of Alceon Corporation **********************************************************