|
|
|
CANCER IN BURNHAM ON SEA NORTH RESULTS OF THE PCAH QUESTIONNAIRE
Chris Busby PhD Helen Rowe BA
Occasional Paper 2002/5 Green Audit: Aberystwyth July 2002
Abstract This paper addresses the results of a unique local cancer prevalence questionnaire survey in the small coastal town of Burnham on Sea, Somerset, near Hinkley Point nuclear power station, undertaken in Spring 2002 and examines the reported cancer risk in Burnham North ward over the periods 1998-2001 and 1996-2001. Volunteers from 'Parents Concerned About Hinkley' (PCAH) called at houses in the ward and gave out questionnaire forms which asked for information about the people living at the addresses and any cancers that had been diagnosed at the address in the last ten years. In addition, the ages, sex and numbers of people living at each address was asked for and this gave a base population which could be used to determine the expected number of cases of different cancer types using England and Wales national incidence data for 1997. 750 completed survey forms were returned defining a base population of about 1500 persons. Results were calculated as Relative Risk or RR which is observed cases divided by the expected numbers based on the national data. Statistical significance was estimated using cumulative Poisson probability. These showed that there was roughly double the breast cancer risk in both periods. (1998-2001: RR = 1.86 10 cases, 5.39 expected; p = .05, 1996-2001: RR = 1.98 16 observed, 8.1 expected, p = .01). This supports earlier work carried out by this group (2000) using mortality data from the office for National Statistics which showed approximately double the national risk. In addition, the data indicated significant excess risks from leukemia (1998-2001: RR=4.1;p = .01; 1996-2001: RR= 2.73, p = .05, 4 observed, 1.46 expected), kidney cancer (1998-2001: RR 4.76 4 observed .84 expected p = .01, 1996-2001: RR = 3.96, 5 observed, 1.26 expected. P = .01), cervical cancer (1998-2001 RR = 5.6, 2 observed 0.36 expected, p = .01; 1996-2001: RR = 5.6, 3 observed, 0.54 expected, p = .01). Reported prevalence of cancer of the colon and prostate were below expectation and for cancer of the lung the prevalence was very low. For all malignanices combined, risk was slightly lower than expected. However, owing to population leakage and early deaths, the true risks are likely to be much higher. On the basis of the year 2001 only, 17 cases were reported with 11 expected, RR =1.55, p = .06. Other questions asked about lifestyle and behaviour in the cancer sufferers. Smoking more than 10 cigarettes per day occurred in 23% of the cancer sufferers compared with 27% in national statistics for all persons and 35% in cancer patients in Irish hospitals in 1996, suggesting that smoking was not a cause of the effect. The most interesting pointer was that 61% of cancer sufferers had hobbies or activities that regularly placed them on the beach or near the sea. The results support the hypothesis that radioactive particles and materials originating from Hinkley Point and other sources which are present in the intertidal sediment have become transferred into the air and become inhaled or ingested by sea coast populations and that the high local dose to tissues is a cause of the effect. Other possible causes cannot be ruled out. The questionnaire study is the first of its kind in the UK and starts a process which will enable communities local to a putative source of risk to bypass the perceived secrecy of the cancer registries and discover the levels of ill health in their neighbourhood.
Background The question of the health effects of radioactive releases from nuclear sites and other sources remains an area of considerable debate. In the last five years there has been new evidence emerging that internal irradiation by the novel substances produced by fission processes e.g. Caesium-137, Plutonium-239, Strontium-90 etc is not adequately addressed by the conventional model. The UK government have recently (July 2001) set up a new committee, CERRIE, to report on the issue. As a part of our research into this area we have previously reported the results of an analysis of cancer mortality data from wards in Somerset in order to examine the effects of proximity to the large offshore mud banks and the tidal regions of the River Parrett which are contaminated by radioactive material from Hinkley Point nuclear power station and also material from other sources such as Sellafield, La Hague, Chernobyl and global weapons fallout. Our results, based on mortality data from the Office for National Statistics (ONS) for 1995-1999 were published in 2000 [Busby et al 2000] and elicited considerable criticism from the Somerset Health Authority (SHA) and the operators of Hinkley Point, British Nuclear Fuels (BNF). Our prior hypothesis was based on a very large study of Irish Sea coastal populations, both in Wales and Ireland, which we conducted in the period 1997-1999 and which had showed the existence of a 'sea coast effect' on cancer. People living near the coast in areas where there were offshore mud banks, estuaries or inlets contaminated by material from Sellafield showed sharp and significant excess risks of cancer of most types which fell off rapidly with distance from the coast or river estuary. Results from Somerset showed the same effect and in particular showed significantly raised breast cancer mortality risk in the town of Burnham on Sea, the largest concentration of population close to the contaminated mud flats. The Standardised Mortality Ratio (SMR) over the period 1995-2000 was about 2, i.e risk of dying from breast cancer was about twice the national average after adjustment for age and social class. The study had been commissioned by Stop Hinkley, and the results were reported in the local media. Somerset Health Authority were dismissive: we had used the wrong populations, they said and the correct populations showed no effects. However, when they supplied us with their correct populations, the effect remained. Later, SHA agreed that there was indeed an excess breast cancer risk but now they argued that it was not caused by the nuclear site and could have occurred by chance alone. They were dismissive and offensive and entirely unhelpful. Recently, we have paid to obtain electronic data and emails written from within SHA on the issue and these have been illuminating, establishing the level of anxiety within SHA and their attempts to minimise the problem. They deny that there is a problem and refuse to discuss the issue. Meanwhile, people continue to die.
Small area cancer data: secrecy and cover-up
In Somerset, as in all areas where we have tried to establish the effects of nuclear pollution, we have asked the authorities for small-area cancer incidence data to analyse. We were only able to examine the Welsh data because we were given the entire database 1974-1989 by the Wales Cancer Registry (WCR) following the orders of the then Director of Health Dr Deirdre Hine to whom we addressed our concerns. This amazing leak resulted in the immediate closure of WCR and its replacement by a new organisation, Wales Cancer Intelligence Unit whose Director, Dr John Steward, moved swiftly to remove large numbers of cancer cases from the inherited database. More than 15% of all the childhood cancer cases recorded by Wales Cancer registry were wiped off the record without explanation. Shortly after this, all the cancer registries in the UK agreed to refuse small area data on the grounds of confidentiality, and ours and other groups requests for such data are now automatically refused. This includes data on incidence produced to the same level of aggregation (e.g. ward level) as the ONS data on mortality(which is available) showing a difference in application of confidentiality considerations which is hard to explain. No amount of pressure will produce the data: it has been refused in Court following a formal request from a judge and has been refused to the Minister of the Environment. The reason given for the secrecy is that it may be possible for someone to be identified if the number of cases of say breast cancer in a ward population of say 1000 women is given out. We are thus asked to believe that the knowledge that there were 6 new cases of breast cancer in Burnham-on-Sea North ward in 1998 would enable us to identify a particular person and for this reason the data must be kept secret. Whatever the reason for this secrecy, it is clearly important to be able to examine the effects of environmental pollution, and as a way around the cancer registries we devised a procedure to obtain the numbers of cancer cases in an area. This simply involved knocking on doors and asking if anyone in the house had been diagnosed with cancer. The method was used for the first time in Carlingford and Greenore on the East Coast of Ireland in 1999 [Busby and Rowe, 2000]. The questionnaire method devised and used successfully in Carlingford enabled us to establish that the sea coast effect in that area was remarkably local with high cancer risk within a few hundred metres of the high water line. Following a meeting in Burnham on Sea in 2000, where these issues were discussed, a local group, Parents Concerned about Hinkley (PCAH) decided to organise a questionnaire-based study of Burnham North ward based on the Carlingford questionnaire. Although this approach has some shortcomings, which will be discussed, it has the great merit of obtaining data has not been filtered through official channels and can therefore be believed, and at base, there will exist a list of people with cancer in a small area and the numbers of cases can be analysed with regard to a base population which is also exactly known. In addition, a questionnaire can ask people about their lifestyle and what they think about the issues involved. This approach breaks through the secrecy surrounding small area cancer risk and has the added advantage of being able to obtain numbers of cancers for the more rare sites e.g. leukemia, where mortality data from ONS is not published or to see if there are any links with activities that might prove epidemiological pointers to the cause of the disease. A version of the Carlingford questionnaire was devised and distributed in Burnham North in the Spring of 2002. It is the results of the analysis of the 750 returned questionnaires that will be reported.
The PCAH/Green Audit questionnaire The questionnaire (see Appendix B) was given by a volunteer from PCAH to each of the houses in Burnham North and the project was explained. It was intended to be filled in for each household on the electoral list by the head of or some responsible person in the household, with the assistance of an interviewer. The questionnaire asked for the sex and age of all persons living at the address. It then asked if any person at that address has been diagnosed with cancer in the previous ten years. Details of this person were then asked for, such as the type of cancer (site), their sex, the age at diagnosis and the year of diagnosis. This data enabled the direct calculation of relative risk in the sample population, relative to the national population. In addition to the fundamental questions above, the questionnaire also asked about the lifestyle of the cancer cases, whether they smoked, whether their habits put them in proximity to the beaches or the sea, whether they ate fish or shellfish regularly and other questions that might throw light on the cause of the cancer In the event, difficulties obtaining enough volunteers and other factors resulted in 749 households returning completed questionnaires. Where there were clear ambiguities in some questionnaires, the house was re-visited. The questionnaires also asked people to indicate whether recipients would be prepared to help further, through interviews or with providing other information. There was also a space on the form for comments.
The population at risk The 1991 census population of Burnham North is given in Table 1. The project was set up assuming a 50% response rate and total ward coverage. In the event about 750 questionnaires were returned which gave a base population-at-risk of approximately 1500 persons. The age breakdown of the respondents is also given in Table 1. It is clear that about one third of the census population of the ward are included in the survey.
Table 1. 1991 census population of Burnham North compared with population included in the PCAH questionnaire survey.
Population leakage The questionnaire asked for details of cancer incidence in the ‘last ten years’. In the event, respondents gave cancer details back as far as the late 1970s. However, it is clear that the inclusion of cancer cases diagnosed in the area for earlier periods runs the risk of missing cancer cases or deaths for any early year because the owner died of cancer (and no one else could report this) or because the house where they lived was sold and new people live there who are not aware of the cancer case and therefore did not report it. This is a problem with such a retrospective study. We call loss of cases from earlier years ‘population leakage’ and it may be examined by looking at the trend in Relative Risk by different period of time prior to the survey. Information may also be obtained by examining Relative Risk by age group since older people are more likely to leak out of the study population because they are more likely to die of cancer and their houses sold. Table 2 gives the prevalence of reported cancer cases by year of reported diagnosis. It is clear that there is an apparent fall off in prevalence with time and this must be due to a population leakage effect. Therefore the apparent risks shown by the data will be less than the true risk. For this reason, the analysis has to consider a trade off between accuracy and statistical power. Two periods of time were used for the analysis, 1998-2001 and 1996-2001. For the six year period, there were 64 reported cases of cancer with 66 expected on the basis of the national average; however, the number of cases for the most recent year was 17 and if this year is representative, there would have been 102 cases in the six years 1996-2001. This latter is more in line with the results obtained by mortality analysis using the ONS national data. Thus the relative risks found on analysis of the questionnaires were adjusted for this effect, they would increase by about 54 percent; however, we cannot reasonably apply such a correction to any particular type of cancer since we do not know the distribution of the leaked (i.e. missing) cases.
Table 2 Cases of cancer reported by year in PCAH survey results
Confidentiality The survey interviewers gave an assurance of confidentiality. Each questionnaire was devised to be returned to the analysis team with the postcode reference only. However, in the event, some were returned with an address. There was space on the questionnaire to ask if the person filling out the questionnaire was prepared to give further help or be interviewed: and a number of people gave their name and address and telephone number. This report will not identify any person or any address and this data will remain confidential.
Method The sample of the population of the study area who were reported in the questionnaire forms were assumed to be the population at risk. The total numbers of cancer cases expected in one year was calculated by multiplying the population at risk in each sex and 5-year age group by the 1997 average national rate for the cancer type or site calculated from the latest (1997) ONS national data [ONS, 2001]. The total numbers expected over 4-year and 6-year period were then obtained by simply multiplying the annual expectation by the appropriate number. This was then compared with the observed number of cases over the same period. The resulting O/E was defined as a Relative Risk and the probability of this result being a chance occurrence was obtained from Cumulative Poisson statistical tables. This p-value represents the probability of the number of cases observed, or less, being found by chance alone given an expectation based on the population in the sample and the national rates. There was no adjustment for Social Class, as this is not possible given the data; however, such an adjustment would not change the risks by more than 10 percent.
Results
The cases are given in Table 3.
|