Results

There were 38 cases (22 boys and 16 girls) of leukaemia recorded between 1978 and 1998 in the population of 0 to 24 year olds in the study region (see table 3 for details). The first case was diagnosed in 1979 and the last in 1998. There were 23 cases of acute lymphoblastic leukaemia (60.5%), eight of acute myeloblastic leukaemia (21.1%), six of chronic myeloid leukaemia (15.8%) and one of unclassified acute leukaemia (2.6%). Five of the cases lived in the Beaumont-Hague electoral ward, 16 in the 10 to 20 km zone and 13 in the 20 to 35 km zone. Three of the cases were Down's syndrome patients, none of whom lived in the Beaumont-Hague electoral ward. None of the 38 cases had a known history of cancer, trisomy 8 syndrome, Bloom syndrome, ataxia telangiectasia, Fanconi anaemia, Li-Fraumeni or immunodeficiency syndromes.

The expected total number of cases in the study area was 36.93, and 38 cases were observed (SIR=1.03; 95%CI: 0.73, 1.41) (table 4). In the 0 to 10 km zone (Beaumont-Hague), there were five cases whereas 2.30 were expected. The SIR for the entire 0 to 24 year old group was 2.17 (95%CI: 0.71, 5.07). The highest SIR for an age group was for the 5 to 9 year olds: there were three cases observed in this group although 0.47 were expected (SIR=6.38; 95%CI: 1.32, 18.65). There was no excess in number of cases for the 10 to 35 km zones for any sex or age group. In the 0 to 24 year old population of Alderney, there were no cases of leukaemia.

Table 5 shows the results classified according to type of leukaemia (acute lymphoblastic leukaemia and acute myeloblastic leukaemia) for children under the age of 15. The SIR was 5.77 for acute lymphoblastic leukaemia in the 5 to 9 year old age group (95%CI: 1.19, 16.86). All three cases involved were boys born in the “département” of La Manche. These cases were diagnosed in 1981, 1989 and 1998.

Discussion

For the area studied as a whole, the number of cases observed was consistent with the number expected. When comparing the three areas around the La Hague plant, the highest SIR of leukaemia among people aged less than 25 years was observed in the 0 to 10 km area (SIR=2.17; 95%CI: 0.71, 5.07). In this zone, the highest SIR concerned the 5 to 9 year olds. These cases were three children aged 5 or 6 and suffering acute lymphoblastic leukaemia.

Of the 194 doctors practising in the area at some time during the study period, nine (4.6%) were dead, four (2.1%) could not be traced, and 23 (11.9%) did not reply. The highest response rates were in the zones less than 10 km and 10 to 20 km from the plant (83.3% and 86.7% respectively). The response rate was 76.9% for the 20 to 35 km zone and 73.7% for the surrounding area, and these relatively low response rates raise the possibility of underascertainment of cases in these areas. It is likely, in any event, that physicians with no cases to report would tend to be found among the non-responders. The files of the usual healthcare system were used, as far as possible, to validate the data, but are not exhaustive for the earlier part of the period under study. All the diagnoses were validated by the same investigator who had no information about the patient (sex, age or place of residence).

In France, general population cancer registers cover the population of a “département”, with the consequence that the population sizes are small. The reference values used were calculated by combining data from registers from five “départements” in which the incidence rates of cancer are stable over time. There are also three registers in France recording paediatric cancer (under 15 years old). According to these paediatric registers, the expected number of cases in the under 15 year old group in our study is 32.42, and 27 were observed (SIR=0.83; 95%CI: 0.55, 1.21). Using these registers for reference, the SIR in the Beaumont-Hague electoral ward in 5 to 9 year old group is smaller (SIR=4.62; 95%CI: 0.95, 13.49; 0.65 expected cases). Because of a more accurate data collection, the paediatric registers give reference incidence rates higher than those from general population registers. Indeed, the expected number of cases of acute lymphoblastic leukaemia in the 5 to 9 year old group calculated from the paediatric registers is higher than the expected numbers in the same age group of all forms of leukaemia calculated from the general population registers.

Linear interpolations between census dates provide an inaccurate estimate of the population because there was substantial migration to and from the study region. During the study period, there were two major building projects in the Nord Cotentin region: the nuclear power station at Flamanville (1978 to 1986) and the construction of a new nuclear waste reprocessing unit on the La Hague site (1982 to 1991). The diagonal interpolation method we used takes account of the aging of the population and of the numbers of births and deaths.15The methodological options taken for both the estimation of the at risk population and the reference incidence rate explained the higher expected number obtained in this study as compared with the previous ones.6 7

Acute lymphoblastic leukaemias are the most prevalent forms of leukaemia among those under 15 years old, and account for the peak in leukaemia incidence observed in young children in industrialised countries.22 In our study, there were four cases of acute lymphoblastic leukaemia in 1 to 6 year olds in Beaumont-Hague, whereas the number expected from the Lorraine paediatric register is 0.96 (SIR=4.17; 95%CI: 1.14, 10.67).20 The incidence of acute myeloblastic leukaemia is more regular during childhood. These observations motivated our detailed analysis of cytological types of leukaemia. In Beaumont-Hague, the excess number of cases of leukaemia seems to be entirely due to the 2.48 excess cases of acute lymphoblastic leukaemia in the 5 to 9 year old group. For the 1 to 6 year old group, the excess was 3.04 over a period of 21 years. For acute myeloblastic leukaemia, the number of cases was insufficient to detect any excess.

Although the La Hague plant began its reprocessing activities in 1966, it was technically impossible to reconstruct the incidence of childhood leukaemia before 1978. A mortality study had been conducted for the years 1968 to 1978.5 For the overall Nord Cotentin area, no excess mortality was observed with no death from leukaemia in the Beaumont-Hague electoral ward. Although mortality is a crude indicator of morbidity for childhood leukaemia, this result may indicate that nothing dramatic occurred around La Hague in the period prior to 1978, when the site was operational.

There have been three previous studies of the incidence of leukaemia in the neighbourhood of La Hague for the periods 1978–90,61978–92,7 and 1978–1996.25 Though these suggested a small excess of childhood leukaemia in comparison with expected, the numbers in each study were very small, and the results therefore somewhat equivocal, leading to the conclusion that it was important that the situation be kept under review. This is the major reason for this study. All cases diagnosed before 1997 (table 3) had already been included in previous analyses. The period 1993–1998 is too short to be analysed separately from the period 1978 to 1992, especially when considering detailed age classes and cytological types of leukaemia. Adding this period confirmed that the incidence of leukaemia in young people in the Beaumont-Hague electoral ward was quite high (SIR=2.17) and showed that it consisted mainly in acute lymphoblastic leukaemia occurring during the childhood incidence peak.

Our data have already been partially analysed, which increases the risk of erroneously concluding that there is an excess. Multiple comparisons performed by age, zone and, cytological type could give rise to at least one significant result by chance alone. In these circumstances, the type I error is not controlled anymore and confidence limits should be regarded only as numerical indices providing guidance for further research. Our findings are consistent with those in similar circumstances: near Dounreay where the SIR is 3.3,26 and at Seascale where the SIR is 10.2.1

Several possibilities have been suggested to explain the excess number of cases of leukaemia observed around nuclear reprocessing plants. They include paternal exposure to radioactivity prior to conception,27 environmental exposure to radioactivity,8 and population movements.28As the results of the case-control study by Gardneret al had never been reproduced, the hypothesis of paternal preconceptional exposure to ionising radiation was dismissed.29 Moreover, the case-control study by Pobel and Viel was not in favour of this hypothesis.8

A study of environmental radiation concluded that it is very unlikely that exposure to radioactive discharge from the nuclear installations in Nord Cotentin could cause a detectable increase in the incidence of leukaemia in Beaumont-Hague30: of the four cases observed from 1978 to 1996, 0.83 could be attributed to ionising radiation, including only 0.0015 due to radioactive discharge from the various nuclear installations in Nord Cotentin. Although the results seemed robust, they correspond to an average estimation and some reserves have been expressed about the lack of an uncertainty analysis. Genetic susceptibility has been evoked as an effect modifier that could potentiate the effect of ionising radiation.31

Dickinson and Parker recently quantified the association between the occurrence of childhood leukaemia and non-Hodgkin's lymphoma, and population mixing among children born in Cumbria.23 The association was all the more stronger when acute lymphoblastic leukaemia among children aged 1 to 6 years were examined. Our study indicates the predominance of acute lymphoblastic leukaemia over childhood leukaemia cases diagnosed in the Beaumont-Hague electoral ward. Although not conclusive, this observation could be linked to the dramatic intensity of population movement that occurred in this particular area. A study is currently underway to analyse population movement and childhood leukaemia in Nord Cotentin. Indeed, it currently seems difficult to dissociate the correlation between the incidence of leukaemia and proximity to the plant from that between leukaemia and population movements.

Several diseases increase the risk of leukaemia.32 33Individuals with Down's syndrome have a 10-fold to 18-fold increased for leukaemia, FAB M7 megacaryoblastic subtype before 3 years and acute lymphoblastic leukaemia after 3 years of age.32 In our study, three patients suffered Down's syndrome and developed acute leukaemia: two of them were acute lymphoblastic leukaemia, the third one could not be characterised. It could have been interesting to produce the analysis without cases suffering Down's syndrome. However, leukaemia reference incidence rates excluding patients with Down's syndrome are not available in France. Moreover, no patient with Down's syndrome was living in the Beaumont-Hague electoral ward when diagnosed. Exclusion of patients with Down's syndrome from the analysis could increase the SIR in this particular area. The 35 other patients presented no clinical or history evidence for trisomy 8 syndrome, Bloom syndrome, ataxia telangiectasia, Fanconi anaemia, Li-Fraumeni or immunodeficiency syndromes. Neither was a pure familial leukaemia observed in these patients.

In view of statistically significant clusters of childhood leukaemia near other European nuclear reprocessing sites, and the concerns of the local population, these findings argue in favour of continued investigations in Nord Cotentin. Three lines should be followed: measuring the incidence of child leukaemia among people who have lived at any time in Beaumont-Hague; identifying possible causes of leukaemia in the area (studies of population movements); and measuring the incidence of other diseases the occurrence of which could be linked to radiation (other cancers, and reproductive function disorders).