Study design is questionable

Editor—We were surprised by the strength of Dominique Pobel and Jean-François Viel's conclusion that “there is some convincing evidence in childhood leukaemia of a causal role for environmental radiation exposure from recreational activities on beaches.”1

The study region was restricted to a circle (diameter 35 km) around the La Hague reprocessing plant. Since their original aim was to investigate the apparent increase in incidence of leukaemia in the vicinity of the plant,2 the design of the study was not ideal for examining recreational activities and food consumption. Cases and controls were tightly matched on geographical region at diagnosis and loosely matched for age. The exposure of interest, parents' recall of their own and their children's dietary and recreational behaviour, may have occurred 30 years or more in the past. Further, there was little information about the control families who refused to participate.

Of the 27 patients ascertained during the 16 year study period, six were young adults (not children) and two had chronic myeloid leukaemia. Given the wide age range, the different diagnostic groups, the small size of the sample, and the multiple tests performed (173 exposure items with one or more methods of testing), we believe that these findings warrant a far more cautious interpretation. Further, the findings rely heavily on tests for trend, which seem predominantly to have been applied even when associations were not statistically significant. The trend statistic is not a traditional epidemiological tool and requires justification under these circumstances. As an example, the figure gives 95% confidence intervals of the relative risk estimates for offspring's recreational activity on local beaches (from Pobel and Viel's table 4). The lower limits are all below 1, and other dose-response relations, other than a positive linear trend, could fit within the range of each estimate.

FIG 1

Risk of leukaemia in association with recreational activity on local beaches

• Download figure
• Open in new tab
• Download powerpoint

Although the possibility that novel pathways of the type proposed should not be discounted,3 we are not persuaded that Pobel and Viel's study provides convincing evidence one way or the other.

Scientific context is needed

Editor—Dominique Pobel and Jean-François Viel found associations between childhood leukaemia and the use of beaches around the La Hague nuclear fuel reprocessing plant in Normandy by pregnant women and children and the consumption of local seafood by children, which they relate to possible exposure to environmental radioactivity.1 They do not, however, pay sufficient attention to the scientific context of their results.

In an equivalent case-control study conducted around Sellafield, west Cumbria, Gardner et al found no associations between childhood leukaemia and these lifestyle factors when using the appropriate “local” control group, and they drew attention to the “low quality” of data based entirely on questionnaires because of the real (and unquantifiable) possibility of recall bias.2 This must be particularly so for births over a period of almost 40 years, as in the La Hague study.

Further, Pobel and Viel do not refer to the vast amount of relevant research carried out on environmental radioactivity over the past decade, which has been summarised recently by reports from the National Radiological Protection Board and the Committee on Medical Aspects of Radiation in the Environment (COMARE).3 COMARE concluded that radiation doses due to discharges are “far too small” to account for the excess of childhood leukaemia near Sellafield. Although Pobel and Viel note the results of the case-control study conducted around Dounreay in Caithness, in which an association between leukaemia and the use of local beaches by children (again based solely on questionnaire data) was also found, they do not refer to the subsequent detailed investigation in the Dounreay area by Watson and Sumner, who carried out a sophisticated programme of measurements of radioactivity in children who had been diagnosed with leukaemia and in unaffected children.4 They found no indication of unusual levels of radioactivity.

Set against the weak and inconsistent epidemiological evidence for an effect of environmental radioactivity on childhood leukaemia in the vicinity of reprocessing plants is the evidence for the influence of unusual population mixing in these areas, of which Pobel and Viel make no mention. Kinlen has produced compelling evidence from a series of studies for the impact on childhood leukaemia of the mixing of urban and rural populations.5 Indeed, given this evidence, it would be surprising if the large construction projects at La Hague and at nearby sites had not raised the risk of childhood leukaemia in this area. The potentially important effect of population mixing on the communities surrounding La Hague should at least have been noted.

Case-control studies have been done in Britain

Editor—As authors of a case-control study of leukaemia in children living near nuclear establishments in southern England, we are interested in the case-control study of leukaemia in young people near the La Hague nuclear reprocessing plant.1 Our study found no links with environmental factors, despite larger numbers and a comprehensive structured questionnaire.2 In our original report of leukaemia in children age 0-14 years diagnosed between 1972-1985, the excess was most marked in children aged 0-4 years living within 10 km of the nuclear establishments at Aldermaston and Burghfield,3 where a doubling in risk was found.

Anxieties have been raised again in the local community by media reports of uranium contamination at the former Greenham Common airbase and a newspaper report of increased mortality from childhood leukaemia at Newbury. In view of this we have collated our clinic figures for a further 11 years, giving data for the whole of west Berkshire for 25 years (table 1). The findings for the two periods are similar, although data for 1986-96 may be incomplete since some children resident in the study area may not have been treated at our clinic. There was no increase in the relapse rate or death rate in those with acute lymphoblastic leukaemia, who have been entered into Medical Research Council trials since 1972 (S Richards, personal communication).

Further studies are planned with the Childhood Cancer Research Group, the Oxford Cancer Intelligence Unit, the Newbury Leukaemia Study Group, and the Berkshire Health Authority. We will assess whether the additional and original cases are geographically related to Greenham or the Atomic Weapons Establishments plants and extend the age range studied to 25. It would be interesting to re-examine the incidence of childhood cancer in the same areas of west Berkshire, as a previous report suggests that the incidence of solid tumours is also raised.4 We hope that some reason or environmental cause can then be postulated.

Deformed insects have been found near nuclear plants

Editor—Dominique Pobel and Jean-François Viel's article touches on people's worries about radiation.1 Before starting to study deformations induced by radiation I worked as a scientific illustrator for the University of Zurich for 25 years, mainly for geneticists and taxonomists, and I had painted leafbugs (Heteroptera) for 15 years in my spare time. Painting mutant fruitflies (Drosophilidae) and normal fruitflies of different families was my main task at the institute and gave me an idea of how nature could look if we go on polluting our environment as we do.

After Chernobyl, I went to contaminated places in Sweden and the Ticino area of Switzerland to look at leafbugs. In summer 1987 I started to collect leafbugs in the areas of Osterfarnebo and Gavle in Sweden and Mendrisiotto in Switzerland, and brought back two pairs of Drosophila melanogaster, which I bred in my kitchen as was done in the laboratory. From the first generation I found many deformations of the eyes, antennae, vibrissae, bristles, segments of the abdomen, and wings. The leafbugs were mostly deformed on their antennae, wings, and legs. In Sweden I also found plants with leaf colour changed from green to dark red or changed colour of flowers; I found heavily deformed leaves in Sweden and in the Ticino.

In 1988 I started to collect leafbugs in the main wind directions of the nuclear power plants Gosgen and Leibstadt in Switzerland, where I found the most worrying changes (fig 1). I continued these studies in Sellafield in 1989, Chernobyl in 1990, Three Mile Island in 1992, and Krummel in 1995.

FIG 1

Leafbugs found in the grounds of the Paul Scherrer Institute, Villigen, Switzerland, an institute for nuclear research; painted in 1990-1

CORNELIA HESSE-HONEGGER

• Download figure
• Open in new tab
• Download powerpoint

Since 1993 I have been studying leafbugs in the canton of Aargau, Switzerland, which contains four nuclear power plants and one research plant. I collect 65 leafbugs at each of 40 points. The insects with the greatest deformities are found close (5-10 km) to the plants, to the east, southeast, and south; less severe, but frequent, deformations are found southwest of the plants.

In my work, I document not just the quantity but also, by painting the deformed leafbugs, the quality of the deformations. When I have published my work in Switzerland, suggesting that there is a connection between low level radiation from the nuclear plants and the malformations, I have been heavily attacked by scientists.

Author's reply

Editor—Jacqueline Clavel and Denis Hémon seem concerned that, as the control selection process relied on general practitioners, there may have been a selection bias. However, the control children were not ill defined, since any parents of a child fulfilling the matching criteria, and not children consulting general practitioners, were asked to join the study. The assumptions about general practitioners' practices are broadly true in this rural area, but I recognise that the stability of catchment areas across time cannot be checked.

Although I have already addressed this issue,1 they still want to know if the controls live farther from the beaches than cases, but this time conditionally on matching. The results remain non-significant for the closest coast, Vauville beach, and Urville-Nacqueville beach (P=0.43, 0.18, and 0.76, respectively).

Among the four significant exposures (use of local beaches by mothers, use of beaches by children, consumption of local fish by children, and children living in a granitic area), only the first two were significantly correlated (P<0.0001); no other pairwise test yielded significant results among controls (0.65>P>0.24). In the electoral ward containing the reprocessing plant, all of the four case children played on the beach at least once a month, whereas only 14 of the 33 controls did so (P=0.05). The independent role played by each of the three children's factors is confirmed by a multivariate analysis despite a lack of statistical power on this small dataset (odds ratios: fish consumption=7.01, P=0.05; beach activity=2.56, P=0.09; granitic area=4.15, P=0.05).

Graham Law and Eve Roman have commented on the parents' recall. Exposures may have occurred far in the past and, as in almost every study, some degree of misclassification is inevitable. But the reliability of maternally reported information, in a similar context, has recently been shown to be affected little by time from birth or determination of case-control status.2 Moreover, media coverage on leukaemia clusters around nuclear facilities had never occurred in France before, nor had a possibility of marine contamination been put forward. Hence, misclassification would be random with regard to case status and should tend to obscure rather than create associations.

Richard Wakeford sets his hopes on the viral hypothesis. Even if a candidate virus is still to be found, and if the British evidence is not so compelling, this is at variance with some French findings.3

I hope this additional information will convince the authors that selection or recall biases are unlikely, and that new methods for identifying the environmental pathways are warranted.