Environmental Decisionmaking

Problem Set 10

1) Assume that in the past ten years an anomalously large fraction of the population of Bugtown has died from a rare liver disease (Buggtown has about 100,000 residents, and an average of 3 per year have died over the last five years). Many in the town suspect that the culprit is an understudied chemical, EZ-kil, found in water leaching from old storage tanks at what used to be a pesticide production facility. The concentration of the chemical found in drinking water in the town is 7 micrograms per liter. People generally drink about 1 liter of tap water per day, and there are no other known risk pathways for the chemical.

There is only one study that has been done on the risk of liver disease from EZ-kil. Researchers took four groups of rats (50 per group, for 200 total) and fed them various amounts of the chemical, and then looked for signs of the disease. In the group that got no EZ-kil, there was no disease observed. In the group of rats fed 1000 micrograms per day, two died from the disease. In the group fed 2000 microrgrams per day, fifteen died from it, and twenty-two died in the group fed 4000 micrograms per day (assume that any rats which got the disease died from it).

Using the results of this study, what is the risk of this disease for a tap-water drinking citizen of Miningtown? (To convert from rat dose to human dose, use the equation: (Human body weight / rat body weight)^ 0.75. Rats weigh about half a kilogram each).

a) (2 pts) Construct a dose-response graph for the rat study. Clearly indicate units on the axes.

b) (1 pts) Fitting a line to the data, what is the equation for the relationship between the dose and the response?

Y=0.000114 (disease rat day/micrgram) * X

Many people left this as 0.0001E-4 because that is how Excel gave it to you. It is important never to trust round numbers unless you are sure of them! In this case, because Excel's default is poorly chosen, your answer will be off by something like 15%. That's actually pretty significant, and it can be as high as almost 100% if you are unlucky (if your answer is 0.000053, and it rounds to 0.0001, for example). To avoid this, always have Excel report values in scientific notation. To do this, simply double-click on the formula, select the "Number" tab, and "Scientific".

c) (2 pts) What is the dose-response equation for humans?

Divide the 0.000114 by (70/0.5)^0.75. In other words:

Y (human) = 0.000114 (disease rat day/microgram) / 40.7 (rats/human) * X
Y = 2.801E-6 (disease human day/microgram) * X

where X is the dose received by humans.

d) (2 pts) Does this chemical seem to be a likely culprit?

Humans are receiving about 7 micrograms per day (since there are 7 micrograms in each liter of drinking water). This is X. So, the risk per person is:

Y = 2.801E-6 (disease human day/microgram * 7 microgram/ (human day) = 1.96E-5 diseases.

In a city of 100,000 residents, you would expect:

1.96E-5 disease/resident * 100,000 residents = 1.96 diseases from the given dose.

Hence, the amount of the chemical in the water supply seems to be sufficient to cause some incidents of the liver disease, and it would definitely be worthwhile to investigate the issue further. Other information we would want to have would be how many cases have been observed, and how rare is "rare".

2) (2 pts) Explain why a linear fit to experimental data may overestimate the actual risk of a certain chemical which is present in the environment only in very low doses.

If there is a "threshold effect" where it takes a certain minimum exposure to cause damage, then the risk at low doses will be essentially nil, but a linear fit will indicate a higher danger. This is because the linear fit doesn't drop as quickly as the response at low doses. Threshold effects can occur, for example, if the body is able to wash out a certain amount of a toxin, but after a given level the defenses are overwhelmed.

3) (2 pt) Can you think of reasons why it might underestimate the actual risk? (You will not get credit for "No", even if it is a verifiably correct response).

One reason might be that the effect saturates after a certain dosage. By this I mean that at a certain level increasing the dose doesn't increase the effect at all. Maybe this is because all of the damage that can be done has been done, for example. Or maybe after a given level the body adapts to the new environmental insult, but it doesn't adapt to lower levels. In either of these cases, a measurement taken at a high level of the toxin would be expected to underpredict the danger of low doses. In class we also showed an example where the danger is not from the chemical itself, but from one of its metabolites. In this case, the body might only be able to metabolize a certain amount of a chemical, and so after that no more damage is done.

4) (2 pts) List some reasons why we use animal models in risk assessment, and why they might not be good estimators of human risk in some cases.

We can't ethically use humans, but we can't model all of the different interactions in the body, so animals are a good compromise. Also, we can carefully control the conditions under which a lab animal is exposed to different toxins, and so we can test various hypotheses upon them. Finally, we can expose animals to much larger quantities of a toxin than humans will be exposed to in the natural environment, so we can get more conclusive results more quickly than we could with epidemiological studies.

On the other hand, there are a number of ways in which animals are different than humans. They have somewhat different physiologies, so it is possible that a toxin acts on a control point that is present in either animals or humans, but not in the other. If it is present only in the animal, then we might overregulate the toxin, considering it to be a danger when it isn't, and if it is present only in humans than we might underregulate the toxin, assuming it is safer than it really is. Laboratory animals are particularly prone to a number of diseases and cancers, and so it is sometimes difficult to distinguish between disease caused by the toxin and disease caused by genetic susceptibility. Animals have shorter lifespans than humans, so some toxins which cause damage that builds up over a long period of time will not necessarily be flagged in animal studies. Also, it is difficult to convert accurately from animal dose/response to human dose/response (the formula given in class is only a very rough estimate, in some cases overestimating the effect and in others underestimating it).

5) (7 pts) In her first major decision as director of the EPA, Christine Todd Whitman overturned the new national standard on arsenic in drinking water (it had recently been approved by the previous director, Carol Brower, after much study). Whitman claimed that the new standard of 10 ppm was too expensive and that we needed to do more research to determine what an appropriate standard would be. Other options were 50 ppm and 5 ppm. Eventually, the EPA decided that changing to the 10 ppm standard was the best course. Assume you are a congressional aid, and your boss wants to know: did Whitman do the right thing?

To answer this question, you will need to determine what level or risk is thought to be associated with each standard? Using that number, roughly how many lives would be saved by changing the standard from 50 to 10? What is the expected cost of the new standards (either total or per household in the most expensive areas)? Does that seem worth it? Can you find an estimate of the uncertainty in the numbers? If you can, does it affect your thinking? If you can't, why do you suppose that is (here you will get no credit for answering "Because I am a poor web surfer", regardless of the veracity of that claim)? Remember after you have done this to go back and answer the original question!

You should be able to find the information you need to answer this quesiton on the web. A good place to start would be the EPA's website (www.epa.gov), in particular on EPA's tecnhical fact sheet on the arsenic issue, but you probably won't be able to find all of the information you need there. If you can't get all of the information, do the best that you can, and explain where you had to make assumptions.

Somehow, the middle paragraph was missing from your problem set. That made it much harder to have any idea what I wanted, and so I graded you accordingly. Still, this should have been an interesting question, so read below to see what I was looking for:

On the web page the EPA indicates that moving from 50 micrograms per liter to 10 micrograms per liter will prevent about 3 bladder cancer deaths per year, and also suggest that the number of lung cancer deaths prevented per year is about the same as the number of bladder cancer deaths. If we assume that human lives are worth about $6 million each, then this would be averting about $36 million per year in damages. The EPA calculates that it will cost about $166 million per year, so from a pure cost-benefit analysis based upon these numbers, it does not appear to be worth the extra money. Notice that EPA gives a range of lung cancer rates "as high as 2-5", which is not particularly helpful because it doesn't give confidence limits or what the low part of the range might be. Uncertainty is difficult to determine, but notice that the NRDC obtained very different numbers in their analysis of a study done by the National Academy of Sciences. Looking at their numbers, we can multiply the approximate total cancer risk at levels between 10 and 50 (the people affected by the rule) by the number of people exposed to that much arsenic, and we find that over 6000 excess cancers are expected without the rule over the lives of the drinkers. Notice that NRDC assumes that people drink 2 liters of water per day (half a gallon), which is probably about twice what people actually drink, so we should probably divide by 2, and then divide by about 75 to get yearly cancer risk, rather than lifetime, so we get 42 excess cancers/year, which is very close to what EPA suggested. The NRDC page gives some sense of the uncertainty in the number of people affected, but not much in the case of risks. These results are largely based on a few epidemiological studies done in Taiwan where very high arsenic levels were found, so you would expect uncertainty to be high. Overall, if you believe in cost-benefit analysis and you believe EPA's national annualized costs of the arsenic rule, then reducing the standard did not make economic sense, based on their data.

Things are actually a little more complicated than this simple estimate, although an analysis of the above would be fine for full credit. In fact, the death rate from lung cancer is 88%, while that from bladder cancer is closer to 25%, so the above analysis underestimates the cost of mortality significantly, since it underestimates the number of lung cancer deaths by over three times. Moreover, we completely ignore the cost of cancer that doesn't result in death. Many people would be willing to pay a large amount to avoid this, although the EPA doesn't know exactly how much. In their final rule (Federal Register, vol 66, no 14, p. 7012) they indicate that they use the willingness to pay to avoid chronic bronchitis as a proxy. This cost is $607,000, according to one study (is WTP the right number to use here? I think so, but why?). These additional considerations lead to a total benefit that is higher than what we calculated, but still slightly lower than the overall estimated cost. EPA also points to numerous additional potential benefits that are not monetized in the comparison, such as the benefits of not fearing arsenic in drinking water (both to well-being, but also to keeping people drinking from tap water rather than using bottled water), and ancillary benefits from installing reverse osmosis systems to remove arsenic (they also remove other things that EPA may regulate in the future).