Classification of 'exposed/non-exposed' subjects in epidemiological studies concerning the possible cancer risk associated with ELF magnetic field exposure is based on the a priori assumption of magnetic field value cut-off points that, often, are defined equal to minimum exposure levels typical of a population residing near high voltage facilities (0.1-4.2 microT), but in some cases an environmental magnetic field level not associated with transmission lines can exist. The results of an ELF magnetic field survey in an Italian urban area (about 1 million inhabitants) are presented: average field levels are correlated with population density of different districts. Exposure indexes are obtained, which are compared with those evaluated in studies regarding domestic exposure: background average levels result in comparable to cut-off points in epidemiological studies, but in some districts with high population density, they are much higher. This shows that knowledge of background magnetic field level in urban areas can assume a significant role in exposure assessment.