Epidemiologist Steve Wing Discusses the Increases in Cancer Rates After Three Mile Island Accident (Part 2)

These talks were recorded live at the Pennsylvania State Capitol on March 26, 2009. Dr. Steven Wing's talk is about the long-term health effects to human, animal, and plant life in the aftermath of the accident at Three Mile Island.

Steven Wing teaches epidemiology at the University of North Carolina-Chapel Hill and conducts research on occupational and environmental health. Since 1988 he has collaborated on epidemiological studies of radiation exposures to workers at U.S. nuclear weapons plants. His 1997 and 2003 articles published in Environmental Health Perspectives describe impacts of radiation from the 1979 nuclear accident at Three Mile Island on cancer rates near the plant. His recent studies examine impacts of industrial animal production and environmental injustice.

Continued From Part 1




SW: I would like to share with you a quotation from one of the affidavits from someone who was near TMI during the accident, because I think it gives us a good example. So this statement was made in 1984, about 5 years after the accident by a man who lived 6 kilometers, about 4 miles, northwest of Three Mile Island at the time of the accident. And he said:

"On Thursday, March 29th, the day after the accident started, I was working all day with my son in the garage. The garage doors were open, so it was warm. People were outside. That night when I took a shower, my face, neck and hands looked like I was at the seashore and got burned real bad. I felt nauseous. My eyes were red and burned. Friday morning when I got out of bed, my lips and nose were blistered and my throat and nose and inside my chest felt like fire. It tasted like burning galvanized steel. My son had similar experiences. He was 22. On Friday, we decided to evacuate. During the next week, I experienced severe diarrhea, which caused rectal bleeding. We took one of our dogs with us, a German Shepherd, female.

Following our arrival in Virginia, the dog passed only blood from the rectum and bled from the nose and mouth. When we returned home the next week, we went in the garage first and found our male German Shepherd had died."

Now elsewhere in this transcript, he describes leaving 50 gallons of water and a very large amount of food for this German Shepherd, the one they could not evacuate with.

"My son and I have both experienced hair loss. Mine was on my head, arms, legs and torso. This hair has regrown. My son lost his hair on his arms and torso, which has also regrown."

Now elsewhere in the interview he speaks about not only were the garage doors open, but they were outside most of the day because it was warm. They were in their shirtsleeves.

So what I began to do is read many, many pieces of testimony. Of course they are all different, but there were several kinds of reports that occurred again and again: the metallic taste, the rashes and burning, the nausea, the hair loss, appear in many of these reports. So how was that explained? Well as I mentioned earlier, people were told that this was a sign that they might have mental instability and it was attributed to stress. And so one of the first things I did when I became involved, interested in this topic was, I undertook a review of the medical literature, and there is one, a quite substantial one on a phenomenon, that is called in medicine, Mass Psychogenic Illness, sometimes in shorthand, Mass Hysteria. So it is clear that it is possible for people to have physical symptoms because of stress. And so I reviewed this literature, looking at the demographic patterns reported in these events and looking at the types of symptoms reported and the patterns that occur in such instances. And what I found was that generally people are within sight of each other, so for example, vomiting is a symptom that is typically reported in these episodes. But the people who vomit, see other people vomit. So it is contagious in the sense of person to person, not in infection, but because of social experience and the meaning of vomiting to us.

But of course here at Three Mile Island in this area, people were not with each other. So it was not a social contagion, as is the typical scenario in these events. And furthermore, I found no instances where people reported hair loss. Rashes are sometimes reported. But the situation did not appear to be like the affidavits I read.

Another thing that is typical in the literature is that there is a predominance of these symptoms among women. Which also was a...of course it is not a census for a systematic random sample of the population. But in the case reports that I read, it did not match up with what is in the medical literature on this topic. But also I was exposed to other information that affected my opinions about what might have happened in 1979 here at Three Mile Island. So one of these documents was a court order that was issued by Judge Rambo, directing Columbia University, which took upon an effort to conduct an epidemiological study of cancer. A court order that directed them to estimate doses to the local population for this epidemiological study. And the court order, and here let me explain that Hatch was one of the Columbia investigators, the court order to the scientists at Columbia University specified that the investigators were prohibited from making “upper limit or worst case estimates of released of radioactivity or population doses unless such estimates could lead to mathematical projection of less than 0.01 health effects.” Now health effects, if that is jargon, it is not really that we’re healthy. It didn't make us healthy. A health effect here means something bad, like a birth defect or a cancer. So the court is basically saying that if the investigators find some sort of worst case scenario that resulted in people saying that basically nothing happened, so 0.01 cases of cancer of course is less than... (inaudible)... So if their calculation comes up with such a small number, yes, you can assume that it is a worst case scenario.

But if it does not come up with such a small number, you can not assume that things were worst case . . . Now what does that mean by worst case? That could be a big range according to what we heard from Arnie Gundersen, worst case. Is it 100 times, a thousand times, ten thousand times worse than what are in the statistics in the official record... I do not know what worst case is. But some people would have a vested interest in deciding what worst case is. And so that is the second part of the order that impressed me, which stipulated that a technical analyst, so this would be some scientist, designated by counsel for the Pools. Now the Pools means the nuclear industry insurers, who are of course, paid for by tax dollars. Concur on the nature and scope of the dosimetry projects, things like what is a worst case scenario? So this court order is specifying that representatives of the nuclear industry get to concur and basically approve what the scientists do in their dose estimates. That seems to me pretty clear that that is not an unencumbered evaluation of the evidence. There is clear potential for influence of the court order and even of the fact that the funding for this work came from industry on the investigators.

So then I was asked to take a look at this study that was done by the group at Columbia, and that is this group Hatch had authored. And so here is a description of some of the technical details of the study. And let me say this: Epidemiology and environmental epidemiology in particular, generally looks at the occurrence of diseases in populations and our goal is to try and see whether there is a connection between some environmental exposure and the occurrence of disease after the exposure.

What we try to do as much as possible, is design a study to be like an experiment. In an experiment, you would assign people to be exposed. Of course in the case of ionizing radiation, it is not ethical to assign people to be exposed, because ionizing radiation is the best known cancer-causing agent. But if it were an experiment, we would take a group of people, say you in this room, and we would randomly pick you, you, you, and so on. And say you will go into this room to be exposed and the others, who were not picked, go into another room, and you are not exposed. And the two rooms will hold everything constant, what air you breathe, what food you eat, everything will be held exactly constant and then sometime later we will evaluate the outcome to see if there is a difference between the two groups. And if there is, it is evidence that it is due to the exposure because everything else was held constant. So that is the idea behind the experiment.

The Columbia investigators actually had a very clever idea in the design of this study. So first they noted that because the accident at TMI was very highly publicized: people in the area knew about it, physicians knew about it. And so they recognized that is a problem if they compared Three Mile Island to an unexposed population somewhere else, there would be no way to avoid this potential bias. So what they did was they identified one study area which is about a 10 mile radius, which at the time had about 160,000 people, and they decided to do the study only within this population. So now they would have to say, what are the exposed and unexposed groups?

So they knew this would be a cancer incidence study, so they went to 19 local and 5 referral hospitals to get information on cancer diagnoses for the period 1975 to 1985. And then they divided this 10 mile area up into 69 small tracts based on the boundaries used by the U.S. census bureau, so that they would know how many people lived there. And then they used the limited monitoring data and release estimates to create a dispersion model to estimate where the radioactive plumes went.

So what it means is that the exposed and the unexposed groups that are being compared are all within the 10 mile area, all exposed to the same news stories about Three Mile Island, the same potential for their physicians to order extra tests, and so on, to avoid this potential for a bias.

So the other thing that is very important to know is that the dosimetrists, the people estimating the doses to the local population, felt that there was too little information to know exactly what the doses were. They just felt they could not say whether it was 1 or 10 or a 100 because of the limited information about the releases. So what they did was rather than give a unit, and Arnie Gundersen talked about “R”, rather than giving a unit of rads or rem or in international units grey or sievert, they simply ordered the 69 tracts from the lowest to the highest, in such a way that if a tract had a value of a 100, they estimate that that is 10 times more radiation dose than a tract with 10. Or if you had 200, that was twice as much as a hundred. So they wanted to get everything on a relative scale.

They also had information on other characteristics of the local population that they used now. They reported these results in the American Journal of Epidemiology in 1990 and they concluded that there was no evidence that gamma radiation doses from the accident had led to an increase in cancer. So it was in the early 1990's that I was asked to review this study and check the calculations.

At first, I told the Aamodts and others that I thought that this study was done, and I believe it was done by a very well-qualified group of epidemiologists, and I did not have any reason to question their work. And I do think the study was very well designed. And let me mention one other thing. Note here that they begin their assembly of cancer records before the accident occurred. And so, this becomes really the most unusual aspect of the design of the study. It is because cancer varies everywhere, regardless of whether they are nuclear ones, for lots of reasons, smoking, diet, and so on, occupational exposures, and variations in background radiation. All could lead to variations in cancer rates. So by collecting information before the accident, having some years of data before the accident, it means we know what the geographic pattern was before this occurred. It would be unfair to assign excess cancers to an area downwind of the plumes, if that area already had high cancer rates before the accident. And that is something that because of the design of the study, because they thought to get the cancer records from the prior years, it is something that we could do in our statistical analysis.

I would note that in the tabular results that they present in their paper, they do not do that. We also noted several other problems with their work. They concentrated mostly on very rare cancers, such as childhood cancers, because it is true that children or the fetus are more sensitive to radiation than adults. But there are so few cases, that it gives little basis for an epidemiological analysis that depends on having had adequate numbers. They also were unable to follow birth cohorts. It is another problem, I will not go into details of that one. They also had some problems, simple data management problems with counting the cases, and also there were incomplete records for 1975 from the hospitals. So they had an undercount of the cases before the accident, because we thought that we should be looking at this pre-accident data and we did not want to have an upward bias saying, well because 1975 is low, it would make it look like the post-accident rates were much higher. So we took out 1975 and re-defined the pre-accident period as 1976 through March, 1979, based on the idea that we did not want that undercount to bias our analysis.

So let me share with you now our findings. They are summarized in this map. Here is the Three Mile Island, the Susquehanna River, Harrisburg...

So these are the estimates from the dosimetrists, under the funding with the court order that I showed previously of where the plumes went. The heaviest plumes are in the dark red to the northwest. And then there are some to the northeast, and a little bit to the southeast, but there are other areas that are very, very low and the numbers (you may not be able to see this in the back) but this green is for zero and the very darkest red goes up some of those tracts. The highest tract is relative to zero, 1,666 units. So there is quite a range according to the scale that they used. And the colors here are tied not only to the estimated doses of radiation, but in the bars, to the lung cancer rates for the time period 1981 through 1985. And the lung cancer rates in this take account of the pre-accident rates. So we are adjusting out whatever the pattern was before the accident and saying, what is the additional amount after the accident?

This line across the middle of the bar said this is the average for the whole 10 mile area after the accident. And you can see that in the green areas with the lowest estimates of the radiation dose, the cancer rates were a half to maybe 80% of the average for the 10 mile area. Whereas, at the highest dose level, the cancer rates were over twice the average for the 10 mile area. And that at each radiation dose category, from zero to the highest, there is an increase, what we would call a linear dose response relationship. So the ordering of the cancer rates after the accident corresponds directly with the estimates of the magnitude of the dose.

Now, what could produce this? It is adjusted for the pre-accident cancer incidence rates. How could we explain this? Well, if you believe that the official estimate of the maximum dose of 100 millirem, less than annual background, is true and we know that with certainty, then we could come up with some other explanations. So for example, and we thought about this quite a bit.

So we’ll suppose people who smoked cigarettes more before the accident, lived in these areas here with low doses during the accident, and they moved after the accident. So when they got their lung cancers, they had moved to these areas. And people here who did not smoke moved from these areas to these other areas. So that kind of pattern is theoretically possible. But we don’t really think that is plausible, that smokers and non-smokers moved from these areas, because nobody of course knew where these plumes were ending up. Even if smokers did have some, if people did have some reason to move according to their smoking habits.

So I believe this is very good evidence that goes along with Arnie Gundersen's suggestion that the releases were potentially a thousand times greater than what has been the story we’ve been told. It goes along with the idea that the symptoms reported by people in the area and the animal deaths were not due to stress, but were due to exposure to radiation, and not just to gamma ray radiation, but potentially to beta and alpha radiation.

And for me, the message for us today as we think about the current plans to expand nuclear power in this country and to promote nuclear reactors, whatever you think about the problems of waste storage or waste transportation or nuclear proliferation or other things, that when we hear, which we do hear often, that no one was harmed at Three Mile Island, that this really should cause us to question that. I would encourage any of you who are interested to really go back and look, as Arnie Gundersen did, into the records, to talk with people who have been here and to respect some of this history because I think it is relative to the debates that we are having today.