Background: Breast-conserving surgery (BCS) has been recommended for m
ost early-stage primary breast cancers, but predictors may vary by tim
e and geographic area. Methods: Among 5,266 early-stage female breast
cancers (diagnosed in 1990-1992) in the population-based Connecticut T
umor Registry, the poverty rate of census tract of residence (an ecolo
gic indicator of socioeconomic status), age at diagnosis, race, marita
l status, extent of disease, year of diagnosis, and town of residence
were analyzed in relation to BCS use. Results: The poverty rate of cen
sus tract was not a statistically significant variable in logistic reg
ression analyses of BCS use; however, age, year of diagnosis, and stag
e at diagnosis were predictors. Residence in a town with a hospital ha
ving radiotherapy facilities or near a university hospital were not pr
edictors of BCS use. High BCS rates (69-94% vs. 49% statewide) were fo
und for residents of a cluster of seven contiguous towns associated wi
th a single (nonuniversity) hospital. Conclusions: BCS was not associa
ted with poverty level of area of residence but continued to be lower
for larger or node-positive cancers. Attitudes and practices of local
physicians were hypothesized as being important in explaining variatio
n in BCS use by town of residence.