Epidemiologic evaluation of the relation between magnetic field exposu
res and cancer depends critically on study design, particularly the me
thods used for exposure assessment. We incorporated a complex magnetic
field exposure assessment protocol into a large incident case-control
study of childhood leukemia. We measured residential magnetic fields
using a standard protocol in current and former homes of 638 cases and
620 controls and determined wire codes for 414 case-control pairs. We
chose a time-weighted average of magnetic field measurements in each
eligible home, weighted by the time the subject lived in each home as
the main exposure metric for each subject. We found that 24-hour bedro
om magnetic field measurements adequately characterize children's resi
dential exposure and that measuring other rooms contributes only sligh
tly to the estimate of average residential exposure to magnetic fields
. Front door measured fields provide useful exposure information when
interior measurements are missing. If feasible, measuring multiple hom
es in which the subject has resided is preferable to measuring a singl
e home. A similar distribution of wire codes for controls agreeing or
refusing to participate in our study implies that risk estimates deriv
ed from wire code data will not be influenced by response bias.