Nuclear weapons test participants study: summary of the fourth analysis
Updated 14 October 2022
© Crown copyright 2022
This publication is licensed under the terms of the Open Government Licence v3.0 except where otherwise stated. To view this licence, visit nationalarchives.gov.uk/doc/open-government-licence/version/3 or write to the Information Policy Team, The National Archives, Kew, London TW9 4DU, or email: psi@nationalarchives.gov.uk.
Where we have identified any third party copyright information you will need to obtain permission from the copyright holders concerned.
This publication is available at https://www.gov.uk/government/publications/nuclear-weapons-test-participants-study/plan-for-the-fourth-analysis-of-the-nuclear-weapons-test-participants-study
Introduction
Three analyses of the cohort of 21,000 veterans and 22,000 control servicemen have previously been performed. Each has examined, in increasingly greater detail, mortality and cancer incidence in the cohort based on following the health of participants. The third analysis was based on follow-up to 1998.
This fourth analysis repeats and extends the previous analyses incorporating information from an additional 19 years of follow-up to the end of 2017.
The results of the analysis are published free to access in the Journal of Radiological Protection.
Objectives of the study
The overall objectives of the analysis remain as defined in the original protocol. Namely to study the impact on mortality and cancer incidence of participation in the atmospheric nuclear weapon test programme.
Similar statistical techniques have been used in each of the analyses but for each analysis as data accumulated it was possible to study the potential lifelong health impact of test participation in greater detail and with greater certainty.
This analysis could not address whether an individual’s disease was caused by their attendance at a test.
It could only examine risks for the participant and control groups as a whole. Furthermore, it could not look at end points other than cancer incidence or mortality, for example arthritis, cataract formation or mental health conditions as data about such diseases is not available for the study.
Statistical methods
Two measures of risk are required to examine the effects of test participation fully. Only by using both together can a reliable picture be formed of the impact of test participation.
Relative risk between participants and controls
The relative risk (RR) measure is the ratio of the rate of a disease in the participants group divided by the corresponding rate among the controls. It allows for changes in disease rates occurring across calendar time, age, rank (a surrogate for social class) and service.
This is the best measure of risk available to use in this study as it compares the participants with the controls who were selected to be similar to the participants, in terms of also being men who served abroad in the same services and in the same type of tropical environment.
Thus, if the RR measure takes a value of one then the risk of a disease in both groups is equal. If it takes a value of 2 then the risk in the participants is double that in the controls.
Standardised mortality and standardised incidence ratios
The RR measure does not tell the full story of any effect of test participation. A RR measure will be raised if there is an excess of disease among the participants or a deficit of disease among the controls or even both at the same time.
Standardised mortality ratio (SMR) and standardised incidence ratio (SIR) are the ratios of the rate of a disease of either the participants or controls compared to the general population of the UK (multiplied by 100) adjusted for calendar period and age.
As all servicemen must pass a physical fitness test at recruitment, we expect even many years later that both groups will have lower rates of disease than the general population – so this is automatically a biased measure; however, we expect it to be similarly biased for both participant and control groups.
Thus in order to best examine the effect of test participation we use the SMR measures to determine if a raised RR is due to excess risk in the participants or a deficit in the controls by looking at the values of the SMR or SIR for each group.
If the SMR or SIR for the participants is appreciably bigger than that for the controls, then we can infer that the RR is raised because of an excess of risk for a disease in the participants. If the SMR or SIR for the controls is bigger, then the raised RR may be attributed to a deficit of risk among the controls rather than an excess among the participants.
P-values
The p-value measures the strength of the evidence provided by the data. A p-value less than 0.05 can be considered to provide good evidence while a value between 0.05 and 0.1 can be considered to provide some but weaker evidence.
These values should not be considered as cut off points as a p-value of 0.049 is not materially different to one of 0.051, rather the p-value should be thought of as providing evidence on a sliding scale between 0 and 1.
Study population
The study population was essentially the same as that in the third analysis but with an additional 19 years of follow-up to the end of 2017 at which time the population had been followed for 65 years.
The average age of those who remained alive was 81 years for both the participants and control groups.
| Status of study population | Test participants (%) | Controls (%) | All men (%) |
|---|---|---|---|
| Alive | 7,301 (34) | 7,718 (35) | 15,018 (34) |
| Dead | 11,906 (56) | 12,547 (56) | 24,453 (56) |
| Emigrated | 1,998 (9) | 1,902 (8) | 3,900 (9) |
| Lost to Follow-up | 152 (0.7) | 143 (0.7) | 295 (0.7) |
| Total | 21,357 | 22,312 | 43,669 |
| Person years | 988,281 | 970,269 | 1,958,550 |
Overall findings for mortality and cancer incidence
The overall mortality rate in the test participants was slightly higher by 2% (RR = 1.02, 90% CI 1.00–1.05, p = 0.04) than that in the control group. This difference was driven by similar increased risks for both all cancers combined (RR 1.03, 90% CI 1.00–1.07) and all non-cancer diseases (RR = 1.02, 90% CI 1.00–1.05).
Variation in background characteristics that could not be accounted for in the analysis (such as smoking habits, diet) is a possible explanation for these small differences.
Both participants and controls had similarly reduced mortality rates compared to the general population (10% and 12% reductions respectively) indicating the likely persistence of the healthy soldier effect.
Leukaemia excluding chronic lymphatic incidence showed evidence of being raised relative to controls (RR = 1.38, 90% CI 1.10–1.75, p = 0.01). Leukaemia risks were driven by increased risks for chronic myeloid leukaemia (CML) (RR = 2.43, 90% CI 1.43–4.13, p = 0.003).
For leukaemia, evidence of increased risk in the early years after the test has generally continued to diminish with time, but for CML, risks have persisted.
Among non-cancer outcomes, only cerebrovascular diseases showed increases in participants relative to controls.
The results of the analysis are published free to access in the Journal of Radiological Protection.