Review of publication bias in studies on publication bias: Here's a proposal for editors that may help reduce publication bias

Childhood cancer and power linesWhat do the data mean?Editor-Draper et al used distance of mother's home from high voltage overhead transmission lines (predominantly 275 kV and 400 kV) at the time of her child's birth as a proxy for her child's subsequent exposure to power-frequency magnetic fields (reviewed by Ahlbom et al). 1 2 As they acknowledge, this is a crude estimate since, in contrast to other reports, 2 no household measurements were taken, no data on more prevalent low voltage distribution sources were collected, no information from other time points in the child's life was obtained, variations during the 33 years period studied were not considered, and no validatory home visits were carried out.A recent report into residential exposures to magnetic fields in the United Kingdom estimated that proximity to high voltage lines, 275 kV and above, explained only 9% of those with measurements ≥ 0.2 microtesla ( T). 3 National data on the distribution of houses in relation to high voltage lines in the UK were provided (J Swanson, National Grid Transco, personal communication, 2000) to the UK Childhood Cancer Study (UKCCS) Group for its study of power lines and childhood cancer, to assess the representativeness of study subjects. 4 An equivalent comparison using National Grid data spanning the far longer period investigated in the paper by Draper et al was not undertaken.The assessments of distance to power lines in the UKCCS were made for all registered controls, who have been shown to represent the general population. 5plot of the distributions of the leukaemia and non-leukaemia cases and controls in the study by Draper et al, national populations, and UKCCS populations by distance from high voltage lines (figure) seems to show that the leukaemia controls used in the analysis presented by Draper et al are systematically different.Their positive result over 100 m may therefore be explained not by an excess of cases but by a deficit of controls in the early years of the study.