Cancer mortality around the Bradwell Nuclear Power Station, Essex

Abstract

Two groups (Green Audit and the Small Area Health Statistics Unit (SAHSU)) have produced reports drawing conflicting conclusions about deaths from cancer, particularly breast and prostate cancer, around Bradwell nuclear power station in Essex. Both groups used mortality data from the same source (the Office for National Statistics (ONS)). There were large differences in the figures presented in the first two reports from the two groups. COMARE asked the ONS to investigate these differences and report back to the Committee. COMARE has subsequently made a detailed study of 3 reports from Green Audit and 2 from SAHSU. All 3 Green Audit reports contained errors in the actual numbers of deaths and erroneous or inappropriate figures for the expected numbers of deaths which, together with inappropriate comparisons of various areas, resulted in over-estimation of the risks. Errors in the first SAHSU report, which underestimated the cancer risks, were corrected in the second. Analyses using correct mortality figures and the most appropriate expected values do not indicate any significant excess of cancer mortality around Bradwell, nor do they indicate any substantial or statistically significant risk of breast cancer mortality in groups of wards bordering the Blackwater estuary, or in Maldon compared with Burnham-on-Crouch.

SUMMARY OF COMARE INVESTIGATION

1. In December 2001 COMARE was asked by the Department of Health to comment on a document produced by Green Audit dated March of that year which concluded that there was a substantial excess of deaths from breast cancer and prostate cancer in the immediate area around the Bradwell nuclear power station. COMARE was also aware of a RIF (Rapid Inquiry Facility) report from the Small Area Health Statistics Unit (SAHSU) to the North Essex Health Authority in May 2001 which was intended to cover the same study population and same time scale as the Green Audit report. There were noticeable differences between some of the figures in the two reports.

2. As both groups had obtained their mortality data from the Office for National Statistics (ONS), we asked our assessor from ONS to investigate the figures and the possible sources of error which must have been present in one or both of these reports. The details of this and further investigations are given in the annex to this statement.

3. In summary, the Green Audit study, based on the simple addition of five sets of annual numbers of the registrations of deaths already aggregated at the electoral ward level, had a net excess of deaths from all malignant cancers of 100 (around 10%) compared with ONS data: results for 39 (three quarters) of the 52 ward totals were incorrect - positive errors totalled 149 deaths, negative errors 49. The total number of breast cancer deaths was only 1 higher than the ONS figure, but results for half of the wards were incorrect: this erroneous redistribution of deaths among the wards contributed to an exaggeration of the risk in four of the seven wards with the alleged largest excess mortality. The total number of deaths from prostate cancer was only 2 below the ONS figure (results for two wards each had a deficit of 1 death). Taking the numbers of deaths from all malignancies in males and in females, breast cancer and prostate cancer together, figures for more than half the ward totals were in error.

4. Green Audit produced a more extensive report on breast cancer mortality around Bradwell in July 2001. Compared with their first report, two of the 13 incorrect observed numbers of deaths had been corrected; and the expected numbers of deaths were higher overall, and consistently across all wards, by around 10%. These expected numbers (unlike the earlier ones) had been adjusted for socio-economic deprivation, but this did not account for the differences from their first report.

5. Checks made against expected numbers of deaths based on updated population figures for the late 1990s and published death rates for England and Wales, indicated that the expected numbers in Green Audit's first report for all malignancies, breast cancer and prostate cancer, and in Green Audit's second report for breast cancer, appear to be low by 12-14% as a result of them using out of date 1991 census populations. The expected numbers of breast cancer deaths in Green Audit's first report appear to be low by a further 10% for reasons unknown, ie by around 25% in total. All Green Audit's standardised mortality ratios (SMRs) are inflated by corresponding amounts, ie by around 14-16% and 30%, respectively.

6. The figures in the SAHSU RIF report to the North Essex Health Authority would not be expected to match the ONS ward data exactly, as they were for actual occurrences, rather than registrations, of deaths (not every death occurring in any one year is registered in that year); and because they were not derived directly from the ward codes, but through the postcode on each individual death record, using various postcode directories. However, the SAHSU totals all had deficits of around 10%: 105 deaths for all malignancies, 11 for breast cancer, and 6 for prostate cancer. The pattern in the deficits at the ward level was quite different from that in the discrepancies in the first Green Audit report: they were concentrated in just five wards where postcodes had been terminated in 1995. COMARE noted that given the rapid turnaround of a few working days, the RIF reports are necessarily based only on data held routinely on the SAHSU database, without any scope for further checking of the data.

7. SAHSU produced a further report for the North Essex Health Authority in March 2002 which satisfactorily explained the reasons for the inaccuracies in its original report. Using corrected figures, the report also demonstrated that that there was no significant excess mortality from all malignancies, breast cancer in females, or prostate cancer, in the area around the Bradwell nuclear power station. In April 2002 the Chairman of COMARE wrote to Dr Chris Busby of Green Audit asking him to comment on the discrepancies between the figures in their first report and ONS figures (which were supplied). Dr Busby was dismissive of COMARE's concerns and unwilling at that time to check the results in the Green Audit report.

8. Green Audit released a third report in December 2002 to the Committee Examining Radiation Risks of Internal Emitters (CERRIE, a working group of COMARE). In it, they explained most of the errors in their first report, and apologised for them. In extending their analysis for breast cancer by two years to 2001, they introduced five further errors resulting in a net deficit of 9 - positive errors totalled 2, negative errors 11 - in the total observed number of deaths; there were several other minor errors in the report. The expected numbers of deaths from breast cancer for 1995-1999 were again the apparently erroneously low figures given in their first report, not the 10% higher figures adjusted for socio-economic deprivation given in their second report. The expected numbers for 1995-2001 were those for 1995-1999 multiplied by 7/5; those for 1995-2001 were therefore also low by around 25% and the SMRs correspondingly inflated by around 30%.

9. In their second and third reports, Green Audit compared breast cancer mortality in various groups of wards around the Blackwater estuary, and in Maldon and Burnham-on-Crouch. There does not seem to be any consistent rationale behind Green Audit's inclusion of wards in these groups. Neither is it clear why Green Audit in their third report, used the expected numbers of breast cancer deaths from their first report, that were not adjusted for socio-economic deprivation, and which were lower than those in their second report. The consequences were to increase the apparent risks. Analyses using correct mortality data and the most appropriate population data indicated no substantial or statistically significant risks of breast cancer mortality in (overlapping) groups of wards bordering the Blackwater estuary, or in Maldon compared with Burnham-on-Crouch.

COMARE is concerned that independent reports using mortality data from the Office for National Statistics could be produced with such noticeably different numbers of deaths. These differences have inevitably led to confusion, on top of the concern among residents in the Bradwell area which was generated by the first Green Audit report. SAHSU subsequently addressed the problem that existed in their first RIF (Rapid Inquiry Facility) report to the North Essex Health Authority and produced an accurate report to the health authority which shows no evidence for excess cancer mortality.

Green Audit has now admitted the errors in the observed numbers of deaths in its first report, most of which were also in its second report. The third report corrects most of these errors but some new errors have been introduced. In addition, all the expected values in these reports are low by 12-14% as a result of the use of out of date 1991 census populations; and there also appear to be errors in the expected numbers of deaths from breast cancer in both the first and third reports (but not the second) which result in them being low by a further 10%, ie by around 25% in total.

COMARE has serious concerns about studies such as those of Green Audit that are published without formal peer review in the standard way that would be carried out by a reputable scientific journal. Such publications often raise a considerable amount of public concern, which is subsequently very difficult to allay if the results are unsubstantiated, as is the case here. Moreover, COMARE wishes to emphasise that any organisation or individual dealing with epidemiological data has a responsibility to ensure that the data are correct before publication.

SAHSU has taken steps to remedy the deficiencies in the postcode directories that led to under-reporting of deaths in its first report to the North Essex Health Authority (which it withdrew). Notwithstanding, the termination and allocation of post-codes is an ongoing process and we recommend that SAHSU ensures that systems are in place to incorporate such changes in their data base at the earliest moment so that their analyses are not compromised.

COMARE notes that the public health officials at the North Essex Health Authority had access to the ONS mortality data and could have checked the figures in the first Green Audit report. Had they done so, the errors in Green Audit's numbers of deaths would have been readily apparent. We recommend that public health officials should, when possible, make their own check on the figures quoted against an authoritative source when allegations of this nature are made.

While recognising that SAHSU's RIF reports to health authorities are produced extremely quickly, COMARE also recommends that should it be decided to give them widespread dissemination, detailed checking of the results should be undertaken and reports should be subjected to peer review. We accept that public concern can cause considerable pressure on public health officials to release unchecked data and also that hindsight makes it easy to criticise such action. Nevertheless, trust is not enhanced by the dissemination of data that later have to be amended.

COMARE also notes that any excess cancer mortality risks at the ward level - whether statistically significant or not - were very small. Owing to i) the small numbers of deaths involved, ii) geographical variation in cancer survival, and iii) particularly for breast cancer, the large number of known risk factors for incidence (on which there was no information at either the individual or aggregate levels), it would not be justifiable, even on the inflated SMRs in Green Audit's reports, to assign causality to any one external agent such as environmental radiation.

Download full statement in Portable Document Format (PDF, 26 pages, 724KB)

Top