Fukushima's disease risk: A major fallout?

Children playing in waves Image copyright AFP
Image caption Nakoso beach, 65km south of Fukushima Daiichi, opened for holidaymaking this week

What claims to be (and indeed appears to be) the first formal attempt to calculate numbers of cancer cases and deaths resulting from the Fukushima nuclear accident has just been published.

The Energy and Environmental Science journal paper calculates that total deaths will lie in the range 15-1,300, while cases will number 24-2,500.

These are deaths among the public, not among workers at the plant.

The calculations were done by Prof Mark Jacobson and Dr John Ten Hoeve from Stanford University in California.

One point they raise in the paper is that while these ranges may be big, even the minima are numbers above zero.

This contradicts statements made since March 2010 that the world's worst nuclear accident after Chernobyl would be likely not to produce a single radiation-linked fatality among the public.

(They also predict a small number of cancers - between two and 12 - among workers at Fukushima Daiichi.)

And the academics say the death and case numbers are lower than they might have been because of luck.

Only one-fifth of the radioactive material vented into the air from the stricken power station fell on land; more fallout on land would have meant a higher casualty list.

As intuition would suggest, the vast majority of the cancer cases would arise in Japan itself.

The estimate hasn't yet made any major headlines around the world; and in part, that's presumably because of the uncertainty ranges.

This kind of exercise is fiendishly difficult to do precisely.

The results depend heavily on what you put into your models, and two kinds of model are needed here - one of how radioactive material dispersed, the second relating exposure to that radioactive material with the subsequent risk of disease.

Image copyright AFP
Image caption Protests against nuclear power continue in Japan, though one reactor has re-opened

Possibly the biggest uncertainty in all this concerns whether there's any added risk of cancer from exposure to low doses of radioactivity.

Theories that are in circulation encompass the idea that there is a risk but it decreases linearly with declining exposure; that there's no extra risk; and that low doses of radioactivity are actually beneficial, a notion known as hormesis.

The Stanford academics used the first of those ideas - what's called a Linear No-Threshold (LNT) model.

One question I raised with Prof Jacobson was whether, using their methodology, you could say what the rates would have been if there were a threshold - if low doses had no effect.

Currently that's not possible, he said, though the figures probably could be derived, given time.

Perhaps the study's most important point materialises when you hold the numbers up to the light of context.

The paper mentions that evacuating people from the power station's environs resulted in as many as 600 deaths - principally the elderly and seriously ill, for whom evacuation was too arduous.

But more context is given by cancer statistics.

The figures given in Prof Jacobson's paper are for lifetime risk - that is, the number of people who will die of cancer at some point.

If you try to make the effect as strong as you can, by assuming that all those deaths will occur in a 10-year period - which is almost certainly too tight - and take the highest end of the range, Fukushima would add 130 deaths per year.

By contrast, more than 350,000 Japanese die of cancer each year, according to the Japan Cancer Society. The disease will cause the death of about one-third of the population.

Fukushima fallout, then, would be adding less than 0.1% to the total cancer mortality. If the real numbers are at the low end of the Stanford range, it'll add less than 0.001%

The nuclear accident has proven a very big deal for Japan and for nuclear power around the world.

Several reasons for that are all too real. Does the added cancer risk amount to another one, from these numbers, or not?