The Taipei Apartment Dweller 9 to 20 year Radiation Exposure

The GKG has been criticized for focusing too much on the radium dial painters. The dial painters' dose rate profiles were internal and alpha particle. Power plant release profiles will be almost all photon and external. GKG uses the establishment's own theory in adjusting for the differences. But now members of the establishment are saying you can't do that. Suppose their right. Suppose their own theory is rubbish as they are now implying. In this case, we need a test of LNT in which the dose rate profiles are external and photon.

Figure 1. Hwang's methodology. Thank you XKCD.

Cobalt-60, Co-60, is a real pain in the butt for reactor designers. All steel contains some ordinary, non-radioactive cobalt, Co-59, as much as 1%. During operation, the cobalt in the reactor pressure vessel and other reactor structure will absorb a neutron and become Co-60. Co-60 emits two nasty, high energy photons (1.17 MeV, 1.33 MeV).

In 2000, a Thai scrap dealer took apart a Cobalt-60 medical source to recover the lead shielding. Four people received acute doses in excess of 5000 mGy.¹ Three of them died. Several others had horrible radiation burns. Co-60's half-life, 5.3 years, is short enough so it emits radioactive particles at a dangerously high rate, but long enough so that you must wait 50 or more years before it's effectively gone away. Cobalt-60 is the reason we let decommissioned reactors sit for decades before we try taking them apart.

In 1982 and 1983, recycled rebar, containing Cobalt-60, was “accidentally” used in the construction of 180 apartment buildings in Taipei. Most of the buildings were completed in 1983. The problem was discovered in the mid-1990's, and full scale investigations started in 1996 after a kindergartener whose classroom was in one of the contaminated buildings died of leukemia.

The Chen 2004 Study

The first study was led by W. L. Chen of the National Yang-Ming University.\cite{chen-2004} The dose rates in the apartments were measured, mostly in 1996. It's not clear how Chen "averaged" his measurements, but we can be sure that the dose rates must have varied significantly depending on where you were in each apartment. One source claims the dose rates were as high as 1 mGy/h, presumably right at a structural wall.\cite{kau-1997}

Since only a single isotope was involved, once you have a 1996 dose rate, it was straightforward to extrapolate back to 1983 and forward to 2002. According to Chen et al, over a period of 9 to 20 years, 10,000 people received an average of 400 mGy each.\cite{chen-2004} Chen divided the exposed population into three groups: High (>15 mGy/y), Medium (5 - 15 ) and Low (<5) the high cohort of 1100 people received a mean cumulative dose of 4000 mGy with a max of 6000. As usual we focus on the high dose group because that's where radiation's effect will show up most clearly. Figure 2 shows the reconstructed dose rate profiles for Chen's most exposed person and his high mean average person, assuming they did not move out in 1996.²

Figure 2. Chen max and high group mean dose rate profiles, no evacuation. The Taipei apartments dose rate profiles are qualitatively and quantitatively similar to the profiles that will be incurred by the public in a power plant release, although the release profile will fall off faster in the first couple of months and decline more slowly after that.

According to LNT, we should have seen about 200 excess cancer deaths in this population. In fact, this population had less cancer than the general Taipei district population with the same age distribution. According to SNT, we should have roughly 1 excess death. Using the profile for the average dose rate in the high group as a proxy for the entire group, under-estimates SNT's over-estimate, but SNT is saying any harm would be undetectable.

Chen argues that the expected number of deaths in his population was 116, and therefore this group is an example of hormesis. This is bogus. In coming up with the expected deaths in the absence of any radiation, Chen uses the cancer mortality for the entire Taiwan population. Cancer mainly kills old folks. These were new, upscale, urban apartments. The residents were a young, upwardly mobile, relatively affluent cohort. Chen made no attempt to control for education or social status or most importantly age. Such basic errors destroy Chen's argument for hormesis. However, as we shall see, the actual deaths in the residents was less than the expected number of deaths in an age matched portion of the population.

The Hwang 2006 Study

In 2006, Hwang et al published a competing study of the Taipei apartment exposure.\cite{hwang-2006} Hwang appears to have done a better job of tracking of the actual residence time of the people in the apartments. One way or another, Hwang ends up with far lower cumulative doses than Chen. Hwang ends up with a mean dose of 47 mGy and a max dose of 2163 mGy. Hwang also excludes a portion of Chen's population because they were unable to determine just when they were living in the apartments. Hwang's collective dose is 299,000 mGy compared with Chen's 4 million. But the bottom line is the same. Hwang's population had significantly less cancer than the general Taipei area population with the same age distribution. The Standardized Incidence Ratio (SIR) is 0.8, -- the cancer mortality rate for the apartment dwellers was 80% that of the control group --- with a 95% confidence interval of [0.7, 1.0].\cite{hwang-2006}[Table III]

According to Hwang, LNT predicts 38 excess cancer deaths, The actual number of excess deaths was -20. Hwang gets the LNT prediction this low by lagging. Under lagging, the dose in the most recent ten/two years for solid/blood cancers is simply ignored, on the grounds that this is the average latency for these cancers. However, in the Kramatorsk exposure leukemia was diagnosed about a year after first exposure.

The Hwang paper is intentionally misleading. The all important abstract, which is often all the media reads, does not even mention the reduction in all cancers, nor the failure of LNT to predict the results. Instead they say

The SIR [Standardized Incidence Rate] were significantly higher for all leukemia except chronic lymphatic leukemia in men, and marginally significant for thyroid cancers in women.

Conclusion: The results suggest that prolonged low dose-rate radiation exposure appears to increase risks of developing certain cancers in specific subgroups of this population in Taiwan.

The first sentence is blatant cherry picking. Hwang broke their results down into 24 different cancers and men and women. The male leukemia statement is based on 6 observed cases when the expected was 2; the female thyroid also on 6 observed cases with 2 expected. In other words, we ignore the overall results and pick through a list of 48 sub-samples until we find two tiny sub-samples that we decide to call attention to.

Hwang's reasoning is nicely illustrated by the famous XKCD jelly bean cartoon, Figure 1. A 95% confidence interval means, if you have 48 samples, the probability that at least one of those samples will show a 95% Confidence Interval is 0.92, even if there is no causal relationship at all. The Taipei apartment data emphatically contradicts LNT. Hwang's methodology in attempting to refute this conclusion suggests that we are dealing with defense lawyers, not scientists.

The Hwang paper was designed to shoot down the Chen paper and reestablish LNT. It was in their interest to minimize the dose rates. We are not told how they came up with the dose rates they used. They do say the measured dose rates varied from 0.5 to 270 uGy/h. A 500-fold variation in dose rate depending in part on where you are in each apartment offers a great deal of latitude in how you convert those measurements into the reported dose rates. Clearly, Chen and Hwang used different methods. I regard Chen's numbers as an upper bound and Hwang's numbers as a lower bound. The truth is somewhere in between.

Hwang tells us very little about his dose distribution, and nothing much at all about the dose rate profiles. ³ We do know his max dose was 2363 mGy. LNT predicts this person alone had a 12% chance of mortal cancer. We don't know how many people had nearly this dose. But we do know that Hwang came up with a 47.8 mGy average for his 6246 subjects, for a total of 298,559 mGy. Hwang tells us 2285 people had less than 1 mGy, and 3157 had between 1 and 50. If we conservatively assume all 2285 got 1 mGy, and the 3157 averaged 25 mGy, then that means the remaining 824 averaged 265 mGy. The LNT mortality for 265 mGy is 0.01325. If LNT were correct, the probability of seeing zero excess cancers for this dose in 824 subjects is (1.0 - 0.01325)**824 or 0.000017. 17 chances in a million.

The Hwang numbers may not demolish LNT as dramatically as the Chen figures; but they clearly meet Feynman's criterion for "one ugly fact", although in this case, the fact that a providential Nature has endowed us with a radiation damage repair system which has no problem with dose rate profiles such as Figure 2 is far from ugly. It's humanity's salvation.

Here's a truly ugly fact. The promoters of LNT can't show us one situation, not one, where people have received a very large dose spread more or less evenly over a protracted period where LNT does not screw up completely. It's quite remarkable that a 100 year old theory that is based on an assumption --- radiation damage is unrepairable --- that we now know is flat wrong, and is always orders of magnitude in error on the kind of dose rate profiles that will be incurred in a nuclear power plant release has survived.

1. Since Co-60's radiation energy is 90% photon and 10% electron, grays and sieverts are numerically equal. I will use grays in this piece to avoid being scolded by the purists for whom sieverts should not exist. For others, feel free to substitute Sv for Gy as you wish.

2. Chen indicates about half the people did move after the kindergartener's death, presumably mostly from the most contaminated apartments. If the max exposed person did move at that time, he would have a taller, thinner dose rate profile, with a peak of just over 2.5 mGy/d.

3. A table showing all the dose rate profiles should be an absolute requirement for any paper that talks about radiation harm to a population. If all we are told is the total doses, the author has already assumed LNT.