**Comments on auto-quiz:
You all got that the advantage of epidemiology is that it "asks the right question." It looks at humans at the test species. But there are several reasons epidemiology cannot answer that question very well for low-level contaminants found in the environment. One is that the human subjects move around, and are not exposed uniformly or consistently. Human populations may be genetically diverse or genetically inbred. Second is that for low level exposure to chemicals reputed to cause relatively rare diseases, it is often impossible to have the statistical power to reach any conclusion. The third is called "confounding factors." Exposure to other chemicals, such as cigarette smoke, is often present. (Because of its potency as a carcinogen and a factor in many other health problems, it is customary to divide all study groups in to smokers and non-smokers, or remove smokers from the study, if that is possible.) A fourth matter the booklet does not mention is that the dose is often unknown. Even for heavier exposures in industry or mass poisonings, the dose is often not known with any certainty and may vary by two or three orders of magnitude.

Regarding the very high doses in animals, several of you did not state the obvious, that it is very expensive to test animals, and the expense of testing the many animals to observe low dose response is impractical.

Questions on uncertainties:
***Q. I have always been taught that the first step in statistical analyses is to check your data for outliers and normalicy. It was kind of treated as an afterthought. It is my belief/understanding that it deserves a little more attention than it seemed to be given in the text.
A. I won't try to defend myself against charges that I am not a statistician. Yes, first step is to look at all your data. But here are some ideas I will throw out: Many types of contamination can vary by many orders of magnitude within a short distance, so an apparent outlier may not be an outlier. Careful with the word "normal." Much environmental data is log-normal. Also, with a large site, there may be pockets of contamination, so the notion of "average" for the site may not give an accurate picture, either.

**Q. When determining risk or site status EPA guidelines do not allow for the use of "non-detects". Instead, protocol is to use 1/2 the detection limit for calcualtions. Also, J-flagged values are considered "usable" because of their "conservative nautre". As someone with more chemical background than most, I have to ask. Why? More specifically, what proof, trends, or justification does the EPA have for using such protocols. Analytical technology increases at an amazing pace every year. We can now detect down to parts per trillion in many instances. Assuming 1/2 the detection limit seems like a beaurocratic hoop not founded in science. Any ideas on this?
A. Yes. It is a "convention" with no real scientific basis. But using "zero" is philosophically wrong, and using the lower limit of detection will give inflated numbers, since most of the samples really are "zero."

*Q. In the discussion about uncertainties, the Hampshire Research Institute mentions that "Even a worker population is typically healthier and more uniform than the average local community exposed to environmental chemicals." Why is that the case? I thought that the opposite is right. The lower average life expectancy of men in comparison to women is often explained with the impact of workplace to men.
A. There are many explanations of women's longevity, including their greater intelligence. Much is explained by their lower rate of tobacco and alcohol use, as well as the traditional reluctance of employer's (in western countries) to place women in dangerous jobs. Monthly iron loss may protect against heart disease. They don't have a prostate gland. Etc., etc. Epidemiology studies always separate the populations by gender. The "healthy worker effect" is one of the commonest "confounding factors" in epidemiology. Workers prone to poor health, or who perhaps developed diseases on the job, leave that job or even the workforce, so only the healthy worker are around for the 20 or 30 years exposure period that may be relevant.

*Q. The Ames and Gold article I found quite informative. I haven't read any on their background. Are you familiar with their work and reputation? Their idea of the non-linear effects of carcinogens seemed well stated. Are you aware of the EPA moving in that direction?
A. Ames and Gold are two of the best known cancer researchers and have been leaders for a long time. Ames is at Berkley and runs an EPA funded research group. The EPA budged just a little, and their standard statement now reads that "the [cancer] risk may be lower than that or may be zero."

*Q. On the various methods of treating petroleum in the risk assessment, I'll have to disagree with you on their applicability. As you probably remember from your work at the Corps, petroleum is on of the major contaminants the Alaska District is cleaning up around the state. We've wrestled for many years on 'how clean is clean?' for petroleum. I believe the indicator-surrogate method is the best we have. And the best we have is adequate (acceptable level of uncertainty) for determining risk and cleanup levels. Although the State of Alaska has regulations that follow this approach (well kind of, sort of), the work done by the State of Massachusetts and the TPH Criteria Working group is well documented and peered reviewed. I think that using the major groupings of GRO, DRO and RRO along with separate analysis of the carcinogens (benzene and the cPAHs) and the TEX (toluene, ethylbenzene and xylenes) gives an accurate picture of the risk. Of course there are uncertainties - !
and they should be stated.
A. Sounds like your up on that topic. You realize that there may be vast differences in the chemical composition of those groups. DRO here could be very different than DRO there. Also, there is a vast difference in their solubility. I'm not saying that such classifications are not a practical answer to the "how clean is clean" question, just that at any particular situation, it may not be correct scientifically. But then, that's true of many chemicals, even when we know the chemical.

Q. Do you think uncertainty is discussed enough with the public when a public agency is making a decision based on statistical data and modeling? There are uncertainties involved in scientific research, which include data gaps that may result from missing specific measurements or from a fundamental lack of understanding about a particular scientific phenomenon. I rarely hear anyone mentioning these uncertainties in any kind of risk forum. What do you think?
A. The wise risk manager, when involving the public, will always mention uncertainties. I do. The problem is that they are…uncertain. A smart antagonist can always point out that you are not sure of this or that. The only defense is to state that, as far as you know, the assumptions are conservative, that is, protective of human health. Also that a "trusted agency," perhaps the EPA, recommended you use those assumptions. That is also a good time to be humble and admit you are just doing the best you can with the site specific data you have and the scientific knowledge you have. This belongs in the next module on risk communications.

Module 10 Index