Statement on the Incidence of Childhood Cancer in Wales
IIntroduction
In August 1998 COMARE were first made aware of a study by Busby et al entitled Proximity to the Irish Sea and Leukaemia incidence at ages 0-4 in Wales from 1974-1989. This study is labelled as the First Report of the Green Audit Irish Sea Research Group and is dated August 1 st 1998. This preliminary study reported that the relative risk of leukaemia in children aged 0-4 years, living in coastal areas of Wales adjacent to the Irish Sea for the years 1974 to 1989, was over four times that expected on the basis of national figures for England and Wales. The authors note that the data analysed in this report were obtained from the Welsh Cancer Registry (WCR) in 1997. The report also notes that this increased level of relative risk persisted for a distance of up to 20 km from the sea and fell off continuously with further distance from the sea through the whole of Wales to its border with England. In the conclusions section of the report the authors state that, in their opinion, the effect is most probably caused by exposure to radioisotopic pollution from the BNFL plant at Sellafield.
At that time we were made aware that the Welsh Cancer Intelligence and Surveillance Unit (WCISU), which had replaced the WCR on 1 April 1997, was concerned that the data used in the Busby et al study did not correspond to that held by WCISU. We were under the impression that the reasons for these differences would soon be identified. However, early in 1999 it became apparent to us that difficulties still remained and we were asked by the Welsh Office to advise on whether the level of childhood cancer in the Welsh Coastal Region gave cause for concern. The Chairman of COMARE, Professor Bridges, contacted Dr Busby who subsequently sent him copies of some of the data supplied to him by WCR, these being data said to pertain to childhood leukaemia for Gwynedd, Mid-Glamorgan and Powys for the years 1974-1989. The COMARE Chairman then organised a comparison of both the Green Audit and WCISU data with the database held by the Childhood Cancer Research Group (CCRG) in Oxford. This group maintains the National Registry of Childhood Tumours, data for which are supplied from a variety of sources including cancer registries but also directly from medical cancer specialists looking after children and as a consequence provides an independent check on data held by cancer registries in Great Britain and is clinically validated. Its ascertainment rate in childhood leukaemia and cancer cases is reckoned to be as complete as is practically possible for such patients.
Data checking process
After carrying out the independent check on the number of cases of childhood leukaemia in these Welsh counties it was immediately apparent that, while the figures for Powys and Mid-Glamorgan were similar in the Green Audit, WCISU and CCRG data sets, there were substantial discrepancies regarding the data from Gwynedd. The data from Green Audit showed 46 cases whereas the CCRG knew of only 8. Subsequently it was established that the CCRG figures for the area considered by Green Audit are similar to those from WCISU but bear no resemblance to the Green Audit figures. In view of this COMARE concludes that the data held by Green Audit, on which the study by Busby et al was based, are incorrect.
Request for advice by COMARE from the Welsh Office
At the COMARE meeting held on 18 March 1999 we were asked by the Welsh Office to examine two papers, the first being the report by Busby et al referred to above and the second being a first draft of a study by Steward et al of WCISU entitled Results of a preliminary study to test the Irish Sea proximity hypothesis of Busby et al. The following is a summary of our discussions.
Analysis of cancer rates
COMARE examined the analysis done by Steward et al to test the Irish Sea proximity hypothesis of Busby et al. This analysis was essentially a repeat of the analysis by Busby et al, but using data from the current WCISU database. We understand that Dr. Busby has indicated his general agreement with the method used in this reanalysis. The analysis finds no evidence for an increased risk of childhood cancer in coastal regions of Wales nor for a decreasing gradient of risk moving inland from the coast, for leukaemia 0-4 years or 0-14 years, brain cancer 0-14 years, or all malignancies 0-14 years (1974-1989). The overall risk for Wales for childhood cancer is not significantly different from that expected from England and Wales rates; the overall relative risk for leukaemia 0-4 years in Wales is 0.82 and in rural Wales 0.79 when compared with England and Wales. We compared the case numbers in the Busby and Steward studies and found substantial discrepancies. We found that the data used by Steward et al concerning the number of cases of childhood cancer, both overall and by specific type of cancer, matched those reported by the CCRG for Wales as a whole, rural and industrial parts of Wales and by geographical band. We conclude, therefore, that the result reported by Busby et al. is due to the use of incorrect cancer registry data.
Data discrepancies
Green Audit have told us that when Busby et al began their work they believed that there were 400 cases in the industrial area of South Wales which they had excluded from their study. However, the authors had subsequently acknowledged that there had been a computer error which had led to the number of cases in industrial South Wales being inaccurately specified as 400 instead of 118. Dr Busby has also informed us that Green Audit has two separate copies of WCR data. He has sent us the first two pages of printouts of these data sets which are dated May 1995 and June 1996. We have been told that the earlier of the two was the one on which the leukaemia analysis was based. The second version apparently contains data for the same period extended to 1990 but it is in a different format. Some small proportion of the data has also been corrupted at the end of the file but we have been told that this does not affect the analysis as the areas are not in the study area. We have been informed that the second set of data has been used as a check against the first for some cancers.
We feel that the need for the original major data revision in the first place and subsequently the finding of an apparently large excess of cases in the study area could have prompted the authors to question the validity of all of the source data with WCR rather than just checking the two data sets against each other. It is good epidemiological practice to make all possible checks on the validity of data to ensure that undue public concern is not raised over an apparent health effect which later cannot be substantiated. Moreover, we would have expected such a discrepancy to be picked up by peer review if the paper had been submitted to an appropriate epidemiological journal.
Nevertheless, the real problem lies with the inability to resolve the confusion in the data used by Green Audit. Dr Busby maintains that the data used are as supplied by the WCR and we have seen no evidence which would cause us to doubt this assertion. Unfortunately it is not possible to identify with certainty the source of the confusion, or even whether it happened before or after the data were received by Green Audit. The WCR kept no copy, either on computer or on paper, of exactly what was released to Green Audit. For this the WCR deserves censure. Failure to keep a record of data released in such a sensitive area represents a departure from the standard of professional competence expected of such an organisation. Following the general re-organisation of cancer registries in England and Wales the responsibility for maintaining such data bases has been passed to WCISU and we are reassured by the fact that their data on childhood leukaemia are similar to those held by the CCRG.
Further investigations
COMARE would like to ascertain exactly how the erroneous data presented by Green Audit arose. To this end the COMARE Chairman has written to Dr Busby to ask him to release to this Committee copies of the electronic files which he received from the Welsh Cancer Registry, so that an attempt can be made to establish the nature and cause of the error. The Chairman has also written to the Welsh Office asking if WCISU can release to Dr Busby data for all cancer sites, by 5 year age groups, for the period 1974-1990, aggregated to the same area of residence level as the WCR data. COMARE anticipates having to issue a further statement following the outcome of investigations into the reasons behind the data errors.
Conclusion
COMARE has found no evidence to support the contention that there is an increased incidence of childhood leukaemia or other childhood cancers among Welsh populations living close to the Irish Sea.
We regard it as unfortunate that premature disclosure of results not subject to peer-reviewed publication has led to needless public anxiety. There is a literature on the management of reports of clusters of diseases, such as the Guidelines for investigating clusters of health events (1990) and the Handbook and guide to the investigations of clusters of diseases (1997), which state that the data on which such a report is based must be verified before publication. We hope that Green Audit will take the earliest opportunity to correct this by withdrawing their conclusion which is clearly not valid. We note that the Welsh Cancer Registry ought to have kept a copy of the data which it sent to Green Audit.
References
1. Proximity to the Irish Sea and Leukaemia incidence at ages 0-4 in Wales from 1974-1989. Busby C, Kocjian B, Mannion E and Scott Cato M. Green Audit Occasional Papers 98/4; August 1998.
2. Results of a preliminary study to test the Irish Sea proximity hypothesis of Busby et al. Steward JA, Adams Jones D, Beer H and John D. Welsh Cancer Intelligence and Surveillance Unit, Cardiff. March 1999.
3. Guidelines for investigating clusters of health events (1990). Morbidity and Mortality Weekly Reports, 39, No RR-11.
4. Handbook and guide to the investigations of clusters of diseases (1997). Information Statistics Division Scotland, Leukaemia Research Fund and Office for National Statistics. Published by the University of Leeds.
24 May 1999
Top
