NRPB- W27: Non-technical summary

Mortality and Cancer Incidence 1952-1998 in UK Participants in the UK Atmospheric Nuclear Weapons Tests and Experimental Programmes

Non-technical summary

This study concludes that overall rates of death and cancer incidence continue to be very similar in participants in UK nuclear weapons tests and in a matched control group. Reports of a raised risk of multiple myeloma amongst test veterans were not substantiated. However, the possibility that test participation caused a small absolute risk of leukaemia other than chronic lymphatic leukaemia cannot be ruled out.

This report describes a third analysis of men who took part in the UK atmospheric nuclear weapons testing programme. The two earlier studies suggested that test participation had no detectable effect on life expectancy or the total risk of cancer. However, the possibility that test participation may have caused a small risk of leukaemia in the early years after the tests could not be completely ruled out.

This third study was guided by an independent Advisory Group. NRPB was originally commissioned to study only multiple myeloma, but the Advisory Group recommended that, just like the other two studies, this should be broadened to all causes of death and cases of cancer in people who are still alive. This analysis extends the follow-up of the previous study by a further eight years - to the end of 1998.

The group of test participants studied here is essentially the same as that for the second analysis. It consists of over 20,000 servicemen and civilians who took part in the tests and a carefully matched control group of a similar size. Controls had served in tropical or sub-tropical areas, but not in test locations. About 85% of those meeting the definition of test participation are included. The statistical precision of the study would be only slightly greater if all test participants could have been included. An epidemiological study does not need to include everyone at risk, so long as people are chosen in an unbiased way. The fact that this study group was essentially fixed many years ago means that possibilities of bias are largely eliminated.

Conversely, had individuals identified because of illness been added to the study, this would certainly have introduced serious bias and invalidated the findings.

Test participants and controls were followed up using the facilities of the National Health Service Central Registers (NHSCRs) which hold records of deaths and registrations of cancer. The staff at the NHSCRs did not know which subjects were test participants and which were controls. Extra checks were carried out by other agencies. A very high degree of follow-up was achieved and the status of test participants and of controls (alive, dead, emigrated, etc) was very similar.

Special checks were carried out on cases of multiple myeloma. These involved comparing the information in this study with data held by the University of Dundee and the Leukaemia Research Fund. These checks confirmed that the data held for the study achieved the high level of completeness indicated by other sources.

Death rates in both test participants and controls were lower than those in men of the same ages in England and Wales for the broad categories of all causes of death combined, for all cancers and for all other diseases. We expect to find lower death rates in working populations and this has been named 'the healthy worker effect' (here 'the healthy soldier effect'). In contrast, death rates for accidents and violence continued to be above the national average in both test participants and controls. The relative risk of death for participants compared to controls was very close to one for each of these broad causes of death.

Up to the end of the previous analysis, mortality from leukaemia (excluding chronic lymphatic leukaemia, CLL, which is not thought to be inducible by radiation) was higher in test participants than in controls. Over the additional eight years of follow-up, rates of both mortality and incidence continued to be higher in participants than controls. The relative risks for test participants and controls continued to be about the same as found in the second analysis, though the increases did not reach statistical significance.

For multiple myeloma, the first analysis had found that the number of deaths in test participants was about the national average, but that there were no deaths at all in controls. This meant that the relative risk in participants was raised. In the second analysis, cases occurred in both participants and controls with rather more in the latter. It was concluded that the increased relative risk found in the first analysis had probably been caused by chance. Over the additional eight years of follow-up in the third analysis, the relative risk was slightly raised for deaths from multiple myeloma but slightly decreased when cases in men who were still alive were added in. Neither the increase nor the decrease approached statistical significance.

The third study also analysed separately special groups who might have been at higher risk from their role in the tests. These groups included those for whom there was a record of exposure to radiation. No striking findings were observed, though the numbers of individuals in some groups were small and hence the statistical uncertainties for rare diseases like leukaemia and multiple myeloma were large.

In summary, overall rates of death and cancer incidence continued to be very similar in test veterans and in a matched control group. Reports of a raised risk of multiple myeloma amongst test veterans were not substantiated. However, the possibility that test participation caused a small absolute risk of leukaemia other than chronic lymphatic leukaemia cannot be ruled out.

What are relative risks?

Relative risks describe the ratio of disease incidence in the participants compared to incidence in the control group. A relative risk of one implies no difference between the participants and the controls. A relative risk of two suggests a doubling, and a relative risk of a half suggests a halving of risk in the participants compared to the controls.

For cancer incidence, relative risks are calculated to include people living with cancer as well as those who have died of the disease. Where more than one type of cancer was mentioned, we made sure that we analysed leukaemia, multiple myeloma or lymphoma in preference to other types of cancer.

The play of chance means that relative risks may differ from one even if the risk to participants and controls is really the same. An increase or decrease in relative risk is often said to be statistically significant if there is only a small chance (one in twenty) of a finding of this magnitude if the risk in participants and controls was really the same.

Last reviewed: 13 October 2009