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? …