Send letter to journal:
Re: Nuclear Test Veterans

Dear Editor

In the March 2003 issue of Occupational and Environmental Medicine, Muirhead et al.[1] described an analysis of mortality and cancer incidence among UK participants in the UK atmospheric nuclear weapons test programme.

Comparisons were made between a pre-defined cohort of test participants and a matched control group. Both groups of men were identified during the 1980s from contemporary records held by the Ministry of Defence (MoD). Comparison with data from other sources indicated that 85% of eligible test participants were included in the study cohort. Mortality and cancer incidence were ascertained in exactly the same way for test participants and controls, using the National Health Service Central Registers (NHSCRs), as is standard for many epidemiological studies in the UK. Only 0.1% of the men were lost to follow-up. Cross- checks were also conducted at other organisations. In the electronic section of the December 2003 issue of the journal, Roff [2] has suggested that the study was flawed because of “under-ascertainment of multiple myeloma”.

This suggestion is incorrect. The critical scientific condition that must be met in a cohort study such as ours is that test participants and controls should be identified in the same way so that the extent of ascertainment of cancers in the two groups will be the same. In this way, the design of the study avoids bias arising through ascertainment of cancers being more complete in the test participant group or in the control group. To supplement cases in an ad hoc way could introduce bias, and we were careful to avoid this. Indeed, it has been suggested that self-response bias occurred in an investigation of US test veterans [3] that used additional ascertainment strategies along the lines suggested by Roff.

Roff refers to the inter-comparison of cases ascertained by NRPB and herself. This inter-comparison is described in more detail in a report [4] published at the same time as the analysis by Muirhead et al.[1] The key points to emerge from the inter-comparison were as follows:

1. Roff did not identify an unexpectedly large number of men falling within the definition of test participants but who were not in the NRPB cohort. Out of 47 confirmed test participants on her list, 9 were not in the NRPB cohort. This percentage, namely 19%, is compatible with the value of 15% estimated previously (two-sided p=0.41). It is not “at least 30%” as Roff appears to suggest.

2. The inter-comparison did not reveal additional death certificates or cancer registrations with multiple myeloma among test participants known to NRPB, during the period for which mortality and cancer data were largely complete. Whilst the completeness of myeloma registration is not known precisely, an exercise conducted for Hodgkin’s disease (5) during the 1970s and 1980s would suggest that about 90% of cases during this period are contained on the NHSCRs. The findings of the inter-comparison and of related checks (e.g. using data held by the Leukaemia Research Fund[4]) are consistent with this estimate.

More generally, Roff has confused the issues of identifying a cohort and ascertaining cases within that cohort. For example, she states that cases not identified by linkage to the NHSCRs were treated as independent responders. This is incorrect: independent responders are men who were identified as being potential test participants on the basis of information other than searches of contemporary MoD records, and whose participation was confirmed subsequently. In particular, many of these independent responders were identified because of their health and hence they represent a selected group. We have presented results from an analysis of mortality among independent responders.[4] However, to add the independent responders to the main cohort of test participants would lead to bias.

We do not accept Ms Roff’s conclusion or the approach that she has taken in her paper. An independent Advisory Group (Chairman: Professor Nicholas Wald) oversaw our analysis and was satisfied that the epidemiological methods we used were sound.

Competing interests: Funding for the maintenance of this database and for its analysis has been provided by the Ministry of Defence.

References

1. Muirhead CR, Bingham D, Haylock RGE, et al. Follow up of mortality and incidence of cancer 1952-1998 in men from the UK who participated in the UK's atmospheric nuclear weapons tests and experimental programmes. Occup Environ Med 2003; 60: 165-72.

2. Roff SR. Under-ascertainment of multiple myeloma among participants in UK atmospheric atomic and nuclear weapons tests. Occup Environ Med 2003; 60: e18.

3. Johnson JC, Thaul S, Page WF, Crawford H. Mortality of Veteran Participants in the CROSSROADS Nuclear Test. Washington, DC: National Academy Press, 1996.

4. Muirhead CR, Bingham D, Haylock RGE, et al. Mortality and cancer incidence 1952-1998 in UK participants in the UK atmospheric nuclear weapons tests and experimental programmes. Chilton, NRPB-W27, 2003. (http://www.nrpb.org/publications/w_series_reports/2003/nrpb_w27.htm)

5. Swerdlow AJ, Douglas AJ, Vaughan Hudson G and Vaughan Hudson B. Completeness of cancer registration in England and Wales: an assessment based on 2,145 patients with Hodgkin’s disease independently registered by the British National Lymphoma Investigation. Br J Cancer 1993; 67: 326- 329.