Circling the LNT Wagons

LNT is the theory that radiation harm is additive. Harm is a function of cumulative dose and nothing else. Exposure period and dose rate are irrelevant. LNT effectively denies our ability to repair radiation damage. LNT has been the foundation of our radiation protection regulation, with tragically pernicious results except for our radiation minders. A rising chorus has called for replacing LNT with a model that does recognize our repair ability. But the Empire is now fighting back. A recently released INWORKs study reasserts the validity of LNT.\cite{richardson-2023} Should we be convinced?

As usual, a properly referenced version is at the GKG web site.

Nuclear Worker Studies

A standard way of testing LNT is to study the health of radiation related workers, often people who have worked in nuclear weapons or at nuclear power plants. These workers are presumed to have elevated dose rates and most importantly they carry badges so the individual dose rates are documented.

A basic problem with this approach is, in almost all cases, the occupational dose rates are tiny. In most cases, they are a fraction of background and medical dose rates. This brings in all kinds of confounding factors. One of the most cited is the Healthy Worker Effect. Most of these studies indicate that nuclear workers have better health outcomes than the general population. This is often attributed to the fact that people with these good jobs have to be healthy to get them, and healthy to keep them. This may be true; or it may not.

The classic example is the 1979-1991 shipyard workers study.\cite{matanoski-1991} This was a case control study in which 28,000 American shipyard workers with nuclear badges were matched with 33,500 shipyard workers doing similar work but not in elevated radiation areas. Both groups were followed for 13 years. The nuclear workers had a mortality rate 25% below the non-nuclear.

But, if it is true, you are saying that the dose rates I'm dealing with are much less important then random fluctuations in health in the general population.

A far more fundamental problem with the nuclear worker studies is they only show us the cumulative doses. We know radiation repair is not only real but is critical to our very existence. This shows up in dose rate dependence. If the momentary dose rate is too high, our repair systems are overwhelmed, and unrepaired DNA damage can lead to cancer.

The radiation workers studies never tell us the dose rate profiles. Even though the dosimeters are read and the doses documented on a regular basis, we are only told the cumulative dose.¹ I know of no exception to this narrow minded focus. The assumption is all you need to know is the cumulative dose. But that's a definition of LNT.² If you make that assumption, you have already assumed LNT is valid. This is as circular as you can get.

Finally, the rad worker studies that assume cumulative dose is all that counts ignore and in effect reject biology. If we have radiation repair systems and if repair takes time--- both established way beyond reasonable dispute --- dose rate is crucial. If the dose rate exceeds the repair rate, unrepaired damage (aka harm) builds up and with it the probability of cancer.

The Fifteen Country Study

The exemplar of radiation worker studies was the 2007 15 Country Study.\cite{cardis-2007} As the authors admit, ``Most studies to date showed little evidence of dose related increase in all cancer mortality". To rectify this unsatisfactory situation, the decision was made to pool all the studies in the hopes that the additional statistical power would reveal the real truth underlying the numbers. But simply adding up a bunch of negative studies was not going to get a positive response. The authors threw out two cohorts, Idaho National Lab and Ontario Hydro, which they admit showed ``strong and statistically significant negative correlation between radiation dose and cancer risk". This was done because the 15 Country Study was stratified by socio-economic status, and the information to perform that stratification was not available from INL and Ontario Hydro.

They also excluded a group of Canadian, U.K, and U.S. workers who had been exposed to substantial amounts of neutron radiation on the grounds that the doses were not adequately measured. This group tended to be both high dose and low cancer mortality rate, 88% that of the included workers. The authors lamely admit ``The reasons for this are unclear and include a possibly stronger healthy worker effect and/or different smoking behavior relative to other radiation workers." What is really unclear is why the dose measurements for this group which were accurate enough for earlier studies are no longer accurate enough to be included.

The case control shipyard study was also not included, presumably because these were not nuclear workers, even though the doses were on average higher than the 15 country average. But strangely a far less controlled cohort from the Portsmouth Naval shipyard was included.

The authors also introduced a set of bias adjustment factors based in part on questionnaires.\cite{thierrychef-2007} This was to correct for differing calibration standards and dose measurement procedures in different countries. In many cases, these adjustments were of the same order of magnitude as the measured dose rates.

At the end of the day, the 15 Country Study was able to come up with a weak and statistically insignificant correlation between cumulative dose and cancer incidence. However the authors admit: ``The study cannot address effects at very low dose rates of the orders of tens of mSv [sic]. Further the power of the study is inadequate to investigate the shape of the dose response, even in the dose range under study."

Even that nearly meaningless result disappeared when Canada withdrew a 3088 worker Canadian cohort which Cardis et al had used, citing problems with the recorded doses.\cite{cnsc-2011} This should have come as no surprise. This cohort of people had an excess relative risk six times that of the 15 Country average. Something had to be fishy. When the Canadian cohort was removed, the excess relative risk for the entire group was not significantly different from zero, even if you circularly assume LNT.

The INWORKS Study

Despite all the manipulation, the 15 nations study had crashed and burned, but Cardis et al were not ready to give up. They continued their study of nuclear workers under the moniker INWORKS.\cite{leuraud-2015,richardson-2015} By this time, they knew their data base pretty well. In particular, they knew that workers in weapons and mixed facilities were showing more harm than workers at power only nuclear plants. They made some adjustments:

1) They excluded the data they had already collected from 12 of the 15 countries because ``of the limited resources and consequent need for efficiency in project coordination." The three countries that survived this project coordination process were France, UK, and the USA. The three countries selected just happened to be the only three countries in the 15 that had nuclear weapons programs. In the authors' words

INWORKS was not intended to assemble the largest number of nuclear workers possible, but rather to assemble those cohorts that were most informative with regard to quality and completeness of exposure and follow up data.\cite{richardson-2015}[p 5]

Once again the American shipyard workers did not qualify.

2) Unlike their earlier work, there is no breakdown between nuclear weapons workers and power plant workers.

After this subjective pruning process, they were left with 308,297 people over the period 1943 to 2005. Figure 1 shows the most recent update of solid cancer. INWORKS only shows one-sided 90% confidence intervals. I've added the more conventional, one-sided 95% CI's. Based on a 95% Confidence Interval, none of the bins are significant. In most journals this would have disqualified them from publication.

Figure 1. INWORKS 2023 solid cancer. Reference \cite{richardson-2023} Based on Supplementary Figure B. The origin is meaningless. The ``zero" dose could easily be 100 mGy for a Rocky Flats worker, and 20 mGy for a Hanford worker.

INWORKS subtracts the local background radiation from the dosimeter readings.³ Almost all the workers were at the very low dose end. The median total occupational dose was 2.1 mGy over decades; but a few people had doses as high as 1217 mGy. This means for almost all their sample, the background dose was far larger than the occupational dose that INWORKS shows us. A median occupational dose, 20 year worker at Rocky Flats, where the background dose is at least 5 mGy/y, would end up in the same bin as a median dose Hanford worker (background about 1 mGy/y) who received 20 mGy over 20 years. INWORKS is not measuring total cumulative dose, which they claim is the all important explanatory variable.

To avoid the Healthy Worker Effect, INWORKS used the 0-5 mGy cohort as their control group. The 5-10 mGy group (less than 0.2 mGy/y occupational), --- a tiny fraction of background --- saw a 6% jump in cancer, relative to the control group. That's a slope of 850% per gray. At this harm rate, the species went extinct a very long time ago. Something else is clearly happening here.

People with near zero occupational dose in nuclear plants have desk jobs. They tend to be higher educated, higher income with different smoking, drinking, and eating habits than the blue collar people. INWORKS ignores these differences, which pretty clearly swamp the effect of an incremental 0.2 mGy/y.

From 5 to 300 mGy, there is essentially no further increase in cancer. So 0.6 mGy to 7.2 mGy over 20+ years causes an impossible jump in cancer; but an additional 250 mGy has nil effect? If this section is linear, it is linear with near-zero slope.

Above 300 mGy, there is a weak but statistically insignificant increase in cancer. But a straight line is only one of the many curves you could put through this data, and one that does not fit it well at all.

But the key question here is: how did these high end people get their dose? Was the dose spread more or less evenly over 20+ years? Or was most of the dose incurred in a handful of sharp spikes? INWORKS refuses to tell us; and they refuse to allow us access to the data. The INWORKS dose data is not available ``for reasons of ethics and permissions from different agencies". Another basic principle of science unapologetically violated. This too should have prevented publication.

But we can be confident that the high end doses were not received evenly. For one thing, that's not how it works. In a properly functioning nuclear power plant, something like 80% of the plant dose is associated with refueling, a two or three week period every two or so years, and even then the doses do not add up to anything like the INWORKS high end. An individual rarely receives more than 5 mGy during a refueling outage. Something very unusual happened to the 0.3% of the people at the high end.

Secondly, if they had received their dose evenly, we would not have seen any increase in cancer. We know this from the Kerala study, Figure 2. People who acquired over 500 mGy over 10 plus years had no higher cancer than their neighbors who received less than 50 mGy over the same period.\cite{nair-2009} And it is likely that most of these people received similar doses rate for most of their lives, not just the ten or so years they were being monitored.

Figure 2. Cancer incidence in 70,000 person Karunagappally cohort.

If INWORKS wants to make the case for LNT, they must rule out dose rate dependence (and with it repair). You cannot rule out dose rate dependence, if you make no attempt to ascertain the dose rates. This is consistent with their circular reasoning. To make the case for LNT, they show us only cumulative doses, which makes sense only if LNT is valid. Round and round we go.

The INWORKS study only raises questions.

1) Why did they reject 12 of the 15 countries' data?

2) Why did they reject the American shipyard study?

3) Why do they show us only occupational doses when they claim the only thing that counts is total cumulative dose?

4) Why don't they use conventional statistics?

5) Why don't they make their data available?

6) How can they argue that dose rate is irrelevant if they don't show us the dose rates?

7) Do the authors really think our bodies do not know how to repair radiation damage?

1. Unfortunately, at most nuclear facilities, the passive dosimeters are read only monthly. But at least the monthly data would point us to any high dose months, which could be followed up by checking the active dosimeter readings for that month. Since the monthly data is computerized, the first step in this process would be trivial.

2. The usual definition of LNT is harm is linear in dose, but that implies harm is additive, which implies the only thing that counts is cumulative dose. And you can go in the other direction. See Replacing LNT with SNT.

3. Passive dosimeters do not turn off when their owners leave the plant. They are measuring the background all the time.