Leukaemia near La Hague nuclear plantBMJ 1997; 314 doi: https://doi.org/10.1136/bmj.314.7093.1553 (Published 24 May 1997) Cite this as: BMJ 1997;314:1553
- Jacqueline Clavel, Researchera,
- Denis Hémon, Heada
- a Department of Epidemiological and Statistical Research on Environment and Health, INSERM, Unit 170, 16 Avenue Paul Vaillant-Couturier, 94807 Villejuif Cedex, France
- b Leukaemia Research Fund Centre for Clinical Epidemiology, University of Leeds, Leeds LS2 9GG
- c British Nuclear Fuels, Risley, Warrington, Cheshire WA3 6AS
- d Royal Berkshire Hospital, Reading, Berkshire, RG1 5AN
- e Feldeggstrasse 21, CH-8008 Zurich, Switzerland
- f Department of Public Health, Faculty of Medicine, 25030 Besançon, France
Bias could have been introduced into study
Editor—Dominique Pobel and Jean-François Viel found an association between going to the beach more than once a month and leukaemia risk, with an odds ratio of 2.87.1 Eighty two out of 192 controls were engaged in this activity. If this association were causal, we could expect, roughly, a twofold increase in the incidence of childhood leukaemia in the Nord-Cotentin. Previously, however, the authors had observed an incidence similar to the expected figures (standardised incidence ratio of 1.1),2 which suggests that a bias could have been introduced in the study design. This issue should be systematically and carefully investigated.
A recall bias could not be ruled out. A selection bias may easily occur since the source population used for identifying controls is ill defined. The control group was identified by the general practitioners who had provided care to the case children. Underlying this design are the tacit assumptions that all the cases were followed by a general practitioner at their time of diagnosis, that those general practitioners are still in practice, and that all the children of the source population were followed by a general practitioner who is still practising. Are these assumptions justified by any evidence? Was the catchment area for these general practitioners identical in 1994-5 (recruitment of controls' parents) and between 1978 and 1993 (recruitment of cases)?
Although the differences are not statistically significant, fathers of cases were ranked in higher social categories than fathers of controls. Some additional criteria would also be useful for comparing cases and controls. For instance, did the controls live farther from the beaches than cases, conditionally on matching? Did they live in similarly sized communities?
To what extent are the variables of interest correlated–mothers' seaside activity during pregnancy, children's seaside activity, consumption of local shellfish, living in a granite house? …