- Dominique Pobel, research epidemiologista,
- Jean-Francois Viel, professor of biostatistics and epidemiologya
- a Faculty of Medicine, Department of Public Health, Biostatistics and Epidemiology Unit, 25030 Besançon, France
- Correspondence to: Dr Viel
- Accepted 28 October 1996
Objective: To investigate the association between childhood leukaemia and established risk factors or other factors related to La Hague nuclear waste reprocessing plant.
Design: Case-control study.
Setting: Area within a 35 km radius of La Hague, Normandy, France.
Subjects: Twenty seven cases of leukaemia diagnosed during the period 1978-93 in people aged under 25 years and 192 controls matched for sex, age, place of birth, and residence at time of diagnosis.
Main outcome measures: Antenatal and postnatal exposure tox rays and viral infections, occupational exposure of parents (particularly ionising radiation), living conditions, lifestyle of parents and children.
Results: Increased trends were found for use of local beaches by mothers and children (P<0.01); relative risks 2.87 (95% confidence intervals 1.05 to 8.72) and 4.49 (1.52 to 15.23) when categories were aggregated in two levels (more or less than once a month). Consumption of local fish and shellfish also showed an increased trend (P 0.01); relative risk 2.66 (0.91 to 9.51) when categories were grouped in two levels (more or less than once a week). A relative risk of 1.18 a year (1.03 to 1.42) was observed for length of residence in a granite-built house or in a granitic area. No association was shown with occupational radiation exposure in parents.
Conclusions: There is some convincing evidence in childhood leukaemia of a causal role for environmental radiation exposure from recreational activities on beaches. New methods for identifying the environmental pathways, focusing on marine ecosystems, are warranted.
The reasons for leukaemia clusters around nuclear installations remain unknown
La Hague in France (the other two are Sellafield and Dounreay in Britain) is one of the three nuclear reprocessing plants operating in the world on an industrial scale, and offers an independent opportunity for research
Some lifestyle risk factors (use of local beaches, consumption of fish and shellfish) were associated with the development of leukaemia among young people
Some convincing evidence was found of a causal role for environmental radiation exposure, whereas no association was found with fathers' occupational exposure to radiation
La Hague (France) is one of the three nuclear reprocessing plants operating in the world on an industrial scale (the other two are Sellafield, England, and Dounreay, Scotland). In recent years there has been considerable scientific and public interest in clusters of leukaemia in children in the vicinity of the British plants, which are still considered as being a matter of concern.12345678 Subsequently, Gardner et al reported that occupational exposure to ionising radiation in fathers before conception of the child yields an eightfold increase in risk of childhood leukaemia and could explain the cluster observed around Seascale.9 These results have been much debated, however, and more recently Doll et al concluded that this association is “largely or wholly a chance finding.”10 Hence, the reasons for the increased incidence of childhood leukaemia around the nuclear reprocessing plants are still largely unknown.
In this respect, La Hague, whose mode of operations and nature of discharges are more similar to those at Sellafield and Dounreay than at other nuclear plants, offers an independent opportunity to shed some light on this issue. Two preliminary studies were inconclusive, but in a recent paper we highlighted a small but increased risk of childhood leukaemia in the electoral ward in which the plant is situated.111213 The aim of our case-control study, which is the first to be carried out in France although heavy investment in nuclear energy has been made there, was to examine whether childhood leukaemia among young people near the La Hague reprocessing plant is associated with established risk factors or with factors related to the plant.
Subjects and methods
The La Hague facility is situated in Normandy (France) in the rural “Nord-Cotentin” area. The study was undertaken within a 35 km radius of the nuclear plant, hence including the usual places of residence of its workforce. Three other nuclear establishments are located nearby, making this study area one of the most densely nuclearised in the world: a contiguous low level radioactive waste depository; a nuclear power station, 16 km away; and the navy dockyards, 19 km away, where submarine nuclear fuel is handled.
The process for the identification of cases has been fully described elsewhere.13 Briefly, a list of cases of leukaemia diagnosed (and histologically confirmed) from January 1978 to December 1993 among people aged under 25 years with a residential address in the study area has been retrospectively compiled from local and regional hospitals and pathology laboratories. Doctors' permission to approach the parents was obtained for all 27 cases.
To circumvent the absence of sampling lists (register of live births, National Health Service central register, family practitioner committee registration, etc) and the French tough regulation constraints (access to nominal data from the population census is strictly forbidden), we have relied on the general practitioners of the area who had delivered care to children with leukaemia. No one had computerised patient records, which could have represented a database for sampling. So general practitioners prospectively identified all adult patients who had a child fulfilling the matching criteria with the index case–that is, sex, age (within 2 years), place of birth (inside the current study area or outside), and place of residence (the same electoral ward or a contiguous one) at time of diagnosis of leukaemia of the corresponding case. If so, doctors filled in a brief descriptive form, whatever the final decision regarding the family participation in the study, and sent it immediately to the research team. To avoid a selection bias quality control procedures were applied. Firstly, doctors were asked to check at the end of each week if they had missed some potential parents of controls among their outpatients and if so to get in touch with them. Secondly, when the identification forms were received, their accuracy was checked by the research team, which declared the control eligible or not. Finally, the research team stopped the recruitment for a given general practitioner when the 10th eligible control was included; but because of tight matching criteria and time constraints on the control identification phase (which ended in October 1995), 235 families were asked to participate in the study. Two refusals, one lost to follow up, five absences of answer despite repeated requests, and 35 ineligible controls who did not meet the matching criteria left 192 eligible controls.
Data were collected from parents by way of face to face interviews at home between November 1993 and January 1996. Parents were asked to sign an informed consent form, and those employed at nuclear establishments were also asked for signed permissions for access to their occupational records. Two trained interviewers administered a detailed structured questionnaire that included information on sociodemographic data, medical history, residential history, lifestyle (recreational activities on beaches, consumption of local fruit and vegetables, drinking of raw milk, drinking well water, exposure to electromagnetic fields), and occupational exposure of parents before the child's conception, during pregnancy, and after the birth. Specific data about antenatal exposures (x rays, viral infections, drug treatments during pregnancy with the index child) were collected. Regarding children, places of residence, lifestyle, viral infections, and x ray exposure up to the date of diagnosis were also recorded.
Radiation dosimetry data for parents ever employed in a nuclear facility were obtained from occupational medical officers who were unaware of the children's status. Radiation details were available in the form of external whole body ionising radiation dosimetry in millisieverts (mSv). We assessed lifetime levels before conception and radiation doses in the three and six months before conception, during pregnancy, and from birth to date of diagnosis of leukaemia. Because of the available data format, these period doses were estimated proportionally from the cumulative doses up to 1982, from yearly doses between 1983 and 1987, and from monthly doses later. The father of one affected child (who had been employed at La Hague only after the birth of his child) and the father of one control refused access to their dosimetries. For a further man hired by a subcontracting company, no information on ionising radiation could be retrieved.
To compare general characteristics between cases and controls, exact Mann-Whitney and χ2 tests were performed with StatXact-3.14 Risk factors analyses were carried out within the sets of cases and controls and findings are represented as relative risks (95% confidence intervals). The results were calculated with the exact conditional logistic regression module of the LogXact computer package.15 Linear trends in the relative risks for ordinal variables were tested by regularly scoring the categories and using these scores as continuous variables. Only univariate analyses were performed, the small numbers eligible for inclusion in the study precluding any powerful multivariate analysis. All the statistical tests were two sided and P values <0.05 were considered significant. For the sake of clarity among the 173 items analysed, only variables associated with a P value <0.20 are detailed except for fathers' dosimetries.
A total of 27 parents of cases and 192 parents of controls were investigated, the median and modal number of controls being 9 and 10 per case, respectively. Table 1 shows the characteristics of cases by subtypes of leukaemia, sex, year of birth, and age at diagnosis. The distribution of various characteristics of parents and children (at time of birth and interview) were similar in case and control groups (table 2).
Neither parents' medical history nor characteristics of pregnancy or birth revealed significant associations (P>0.20, results not shown but available on request to authors). Regarding viral infections during childhood we found a relative risk (95% confidence interval) of 17.95 (0.93 to 1060.49) for glandular fever (two cases and one control).
Table 3 presents some details about maternal lifestyle during pregnancy. The analyses were restricted to the 23 mothers who lived during their pregnancy and gave birth in the study area. Analysis by recreational activities on local beaches indicated a significant increased trend (P<0.01) and a relative risk of 4.49 (1.52 to 15.23) if categories were grouped in two levels (cut off point of once a month). Analyses by eating habits, use of seaweed as fertiliser, exposure to electromagnetic fields, or characteristics of residences did not indicate any associated risk (P>0.20, results not shown).
Table 4 summarises some aspects of the children's lifestyle. Results for recreational activities on local beaches showed a significant increased trend (P 0.01) and a relative risk of 2.87 (1.05 to 8.72) when categories were aggregated in two classes (cut off point once a month). Consumption of local fish and shellfish yielded similar findings with a significant increased trend (P 0.01) and a borderline significant increased relative risk of 2.66 (0.91 to 9.51) if categories were grouped in two levels (cut off point once a week). Restriction of analyses to the 209 children born in the study area gave quite similar results with significant increased trends (P 0.04 for both factors) and relative risks of 2.34 (0.82 to 7.30) and 2.31 (0.71 to 9.89) for use of local beaches and consumption of fish and shellfish (same cut off points as above), respectively. There was no evidence of any relation with drinking local raw milk or use of seaweed as fertiliser (P>0.20, results not shown). We found an increased relative risk of 1.18 (1.03 to 1.42) a year for homes made of granite materials or built on granite ground. Relative risks around unity were observed for various surrogates of exposure to electromagnetic fields or for time lived in homes with double glazing (P>0.20, results not shown).
There was no evidence of any association with mothers' types of occupational exposures in any period (various chemical products, wood dust, radioactive materials, ionising radiations) (P>0.20, results not shown). A few mothers claimed to have been exposed to radiation, but this led to non-significant relative risks whatever the period considered (P>0.20, results not shown). Fathers' types of occupational exposures were also not related to leukaemia in any period (P>0.20, results not shown). A few fathers claimed to have been exposed to radiation, but no trend seemed to be significant whatever the period considered (P>0.20, results not shown).
According to employment and radiation records none of the fathers of children with leukaemia had detectable lifetime doses before the conception of the child, whereas doses for fathers of controls ranged from 0.15 mSv to 79.00 mSv for the whole period, from 0.03 mSv to 9.10 mSv during the six months before conception, and from 0.02 mSv to 4.62 mSv during the three months before conception. During the gestational period, doses ranged from 0.40 mSv to 1.97 mSv for cases and from 0.06 mSv to 13.29 mSv for controls, and during the postnatal period from 0.20 mSv to 26.10 mSv and from 0.30 mSv to 162.45 mSv, respectively. Table 5 presents corresponding relative risks. We found no significant trend whatever the period of exposure.
Our main finding was that some lifestyle factors are associated with the development of leukaemia among young people, suggesting contamination with radiation through an environmental route.
Potential limitations of our survey, as of any case-control study, should be noted. The possibility that a selection bias in the prospective identification of controls has strongly influenced the results does not seem likely for several reasons. Firstly, a two step quality control procedure (internal and external) was used. Secondly, parents and not children were recruited when they visited their general practitioner to avoid the selection of a population having complaints and calling for medical consultations. Thirdly, the high motivation of the general practitioners resulted in a high participation rate (96%) among potential eligible families. Fourthly, cases and controls were comparable for sociodemographic characteristics and perinatal conditions. Although we chose our methods to bypass legal constraints and to cope with restrictive matching criteria, matching controls on general practitioner's catchment area led (more or less) to matching on socioeconomic status and neighbourhood, hence potentially decreasing confounding effects.16
Recall bias can be ruled out, in our opinion, at least for our main findings. Firstly, in contrast with Great Britain, neither public debate nor media coverage on leukaemia clusters around nuclear facilities occurred in France before December 1995 because of previous negative results.11 12 A cluster around La Hague was suggested at that time,13 but marine contamination had never been considered.13 although an aerial route was hypothesised. Secondly, the data collection from cases was completed in March 1995. Hence parents of children with leukaemia would probably not link such exposure as a possible cause of their child's disease. Thirdly, nuclear occupational results are not prone to recall bias because they are derived from dosimetry files. Fourthly, the French public does not yet seem concerned about radon exposure. As a precautionary measure, we did not use the word “radon” in our survey.
Non-significant findings are difficult to interpret because of the wide confidence intervals on many of the relative risks given in relation to the small number of cases included. By performing exact conditional statistical analyses, however, we have used the most powerful statistical tool available. When some data cells are empty, maximum likelihood estimates do not exist whereas the exact method can yield median unbiased estimates, exact confidence intervals, and exact P values. Some chance findings, however are likely because of the number of variables under study.
Our main finding was that the use of local beaches by children and mothers was associated with the development of leukaemia among the children. If ionising radiations are involved, this suggests a totally different pathway to occupational exposure, putting the environmental hypothesis in the limelight again. Supporting evidence for a causal effect comes from various factors. Firstly, a dose-response effect is highlighted. Secondly, a chance finding is unlikely because of the low P value (0.01) and the fact that another significant variable (consumption of local shellfish and fish) also points to a potential marine pathway. Thirdly, confounding by geographical distribution of places of residence (residence closer to the sea could increase the likelihood of a child visiting a beach to play) is not plausible because of the matching of control on general practitioner (and then on neighbourhood). Fourthly, this result does not stand in isolation, being in agreement with the study conducted by Urquhart et al around Dounreay.17 The use of local beaches before diagnosis also appeared as a significant risk factor in this area (P 0.04).
An earlier study has found an excess of childhood leukaemia in the electoral ward containing the reprocessing plant.13 A pertinent question is to what extent this excess may be explained by the demonstrated association with use of beaches. All four children with leukaemia who lived in this ward played on the beach at least once a month compared with 13 out of the 23 similarly affected children who lived in the remaining area (exact P trend on six categories 0.14). Hence, the observed geographical excess can be explained by the association of playing on the beach with leukaemia.
At La Hague most routine releases of radionuclides have been discharged through chimneys. Liquid releases occurred by direct piping below the surface of the Channel. Besides, leaks from the contiguous low level radioactive waste depository to groundwater pathways and releases from the two other nuclear establishments of the study area (both on the coastline) may have contributed to marine radioactive exposure. Nevertheless, in a comparison of the amounts of liquid releases (15 times lower than gaseous releases) the evidence would rather lie in the environmental pathways by which radiation reaches the body rather than in the quantitive aspect of releases. Individual lifestyle (bathing and playing on the beach) can extend the contact with radionuclides in the environment and result in possible uptake of high levels of contamination.18 Besides, larger particles of insoluble materials can settle out near the discharge point. Radionuclides in particulate form are especially subject to concentration by filter feeding organisms and then by the whole aquatic food chain. As a result, intake of radionuclides such as strontium-90, which tends to concentrate in bones, is probably higher for consumers of small fish. Finally, for radiation exposures from external sources in the environment, the absorbed dose to body organs increases with decreasing body size, as with children.18
Our data clearly do not support a genuine association of leukaemia with fathers' occupational exposure to radiation. Among fathers of children with leukaemia, none had accumulated any dose before conceptional, detracting from the conclusions of Gardner et al,9 and no significant trend was seen during the subsequent periods. These results are in line with those from two case-control studies conducted around Dounreay and in Ontario.17 19 Apart from the fact that Gardner's results could be due to chance alone,11 two explanations for the discrepancy between findings from Sellafield and La Hague can be put forward. Firstly, figures for all the control fathers indicated that radiation exposure in workers was somewhat lower in La Hague. For cumulative preconceptional doses, 89.5 % (17/19) lay within the range 1-49 mSv at La Hague and only 60 % (45/75) at Sellafield. Secondly, there may be differences in the occupational exposures of fathers in the two settings (internal exposure to tritium, concomitant exposure to toxic chemicals, etc).
We found some evidence of an increased risk of leukaemia associated with a surrogate for radon exposure. We verified the presence of granite in the ground or in the building materials during the home interview, but unfortunately could not do so for the previous houses inhabited by the families. Nevertheless, the strength of the association reported (relative risk 1.18 a year, P 0.01) deserves consideration. Some studies have suggested a similar connection either with terrestrial γ radiation or radon, but a recent one has challenged this association.20 21 22 23
On the whole, some convincing evidence is found of a causal role for environmental radiation exposure operating through recreational activities on beaches or consumption of fish and shellfish, whereas there is no confirmatory evidence of the association represented by Gardner et al.9 But one explanation probably does not account for all cases, and other exposures such as to radon may play some part, maybe even a synergistic one. New methods for the identification of the environmental pathways (focusing on marine ecosystems) and their integration in the dose reconstruction process for children are clearly warranted.
We thank Dr Carre and Mr Boceno for their efficient cooperation in the field; the hospital physicians who identified the cases: Drs Boutard, Danel, Malet, Barrelier, and Reman; the general practitioners, highly motivated, who introduced the study to their patients: Drs Aerts, Bauzon, Bienvenu, Burtin, Cau, Compagnon, Decourselle, Dejonghe, Dolbecq, Dossmann, Etienne, Gigan, Gras, Haquet, Houly, Huet, Lebon, Lenoel, Lepleux, Lerosier, J Letellier, JR Letellier, Levast, Lucas, Mougenot, Navez, Olivier, Rabay, Rousseau, Servant, Sevin, Stefani, Varin, and Vidal; the occupational medical officers who provided the radiation dosimetries: Drs Ducousso, Fournet, and Royer; and Mrs Tilatti, Mrs Ninucci, and Mr Vincent for their technical support.
Funding: This study received primary financial support from Institut National de la Santé et de la Recherche Médicale (grant 931010) and supplemental support from Ligue contre le Cancer (comité départemental du Doubs), France. Dr Pobel received grants from Ministère de la Recherche and Ligue contre le Cancer (comité départemental de l'Orne), France.
Conflict of interest: None.