The continuing and disturbing releases of radioactivity from the damaged nuclear reactors at the Fukushima Daiichi plant in Japan (see Nature 471, 555–556; 2011) raise questions in everyone's minds. What are the risks for the nuclear plant workers? For the local population? For the rest of Japan? Worldwide?

Although the scientific community is doing what it can to estimate these risks, the reality is that we really don't know. More specifically, the uncertainties associated with our best estimates of the health effects of low-doses of radiation are large1. And not knowing the risks means that we really don't know what is a reasonable evacuation zone, whom to evacuate, when to evacuate or when to allow people back.

In fact, the different short-term evacuation zones that have been established by the Japanese government or recommended by the US Nuclear Regulatory Commission for Fukushima are simply the result of different assumptions about what releases may potentially still occur at the Daiichi power plants. But even if we knew the final extent of the releases and the extent of the population exposures, we do not know enough about the possible effects of low-dose radiation on health to be able to make rational decisions regarding evacuations. We don't know the risks for an 'average' person, and we certainly don't know the risks for more radiation-sensitive populations such as children or individuals with genetically based radiation sensitivity. So the immediate logistic responses to the current situation in Japan are based on little more than educated guesswork.

The immediate logistic responses to the current situation in Japan are based on little more than educated guesswork ,

The uncertainties over the long-term consequences of the releases are even more troubling, because they potentially affect much larger populations than those in or near the immediate evacuation zone. We do not know the long-term significance of the very low levels of radioactivity that will remain for some generations in food, water and the rest of the environment. We don't know this for the tens of thousands of local residents who will receive small extra radiation doses, for the millions of people who will receive very small extra radiation doses, or indeed for the hundreds of millions of people who will receive minuscule extra radiation doses. For almost everyone, any increase in individual cancer risk will be very small indeed, but we do not have a good understanding of the public-health consequences of millions of people all being exposed to minute increases in cancer risk.

How is it that we don't know enough about the biological effects of low doses of radiation, either for individuals or for populations? We have been studying the risk of radiation since the discovery of radioactivity more than a century ago, and quite intensively for more than 50 years2. Surely by now we should know most of what we need to know, to make informed science-driven decisions?

Hard to measure

The short answer is that directly detecting and quantifying effects on health in populations exposed to low doses of radiation is hard — and often impossible. The long-term health effect of most concern associated with small doses of radiation is cancer: but given that about 40% of any group of people will get cancer anyway at some point in their lives, trying to measure a very small radiation-related increase in this cancer rate results in tremendous uncertainties, unless the exposed population is very large and the individual radiation doses are comparatively well known.

One obvious, and largely untapped, source of information is Chernobyl. The 1986 reactor accident in the former Soviet Union was far worse than the current situation in Japan, potentially enabling us to study a 'worst possible' nuclear-accident scenario. Because the exposed population was large and exposed to a wide range of radiation doses, there are significant opportunities for informative population health studies. But 25 years on, with the exception of studies of a couple of specific cancers — thyroid cancer and leukaemia — no large-scale systematic cancer studies have been carried out on the exposed populations (see Nature 471, 562–565; 2011 & 471, 547; 2011). Much of the key dose-estimation work has already been done3, and we need to expand our efforts here to extend these studies to all the common cancers.

We also need to start assessing whether it would be reasonable to undertake large-scale population studies among the exposed populations in Japan. A key factor here, as with any study of radiation risk, will be to establish individualized radiation-dose estimates for every exposed person4.

But there are fundamental limits as to how much information can be extracted from studies of populations exposed to low doses of radiation. The number of radiation-related cancers will always be very small compared with the 'natural' background number of cancers, and we have no way of distinguishing a radiation-related cancer from any other cancer. It follows that population studies alone cannot provide the risk information that we need at very low radiation doses.

Basic science needed

So we also need to take a complementary approach, studying the basic mechanisms by which low doses of radiation cause cancer — at the level of genes, chromosomes, cells and organs. Here progress has been slow, because the mechanisms are very complex. But in the long run, augmenting the results of population studies with an understanding of mechanism is the best way, and arguably the only way, to get the information that we need about the health effects of very low doses of radiation5.

In the United States there is just one programme, at the US Department of Energy, that focuses on supporting the basic science behind low-dose radiation risks — and it faces heavy cuts, or even elimination, in the current round of US budget negotiations.

All these uncertainties make it hard to frame a debate about the future of nuclear power in Western countries. We are clearly at a fork in the road regarding nuclear power, where we will either have to replace our ageing fleet of reactors or move away from nuclear power entirely. To make rational decisions about these momentous questions, we need to understand the risks of low doses of radiation with a great deal more certainty. Otherwise the debate will be framed around the extreme positions of 'radiation is universally dangerous' and 'low doses of radiation pose no risk'.

If we want to make rational decisions in responding to nuclear accidents or radiological terrorist events, or make rational policy decisions about the future of nuclear power, not to mention the rapid increase in medical imaging such as computed tomography (CT) scans, or even the new airport X-ray scanners, we must redouble our efforts to understand the health risks of low doses of radiation.

David J Brenner