Occasional reports have suggested an unfavorable effect of high estrogen receptor (ER) concentrations in primary breast cancer. In a population-based study we identified a subgroup explicitly exhibiting this seemingly paradoxical effect. ER concentrations were prospectively measured in a single laboratory by multipoint DCC assay. The relative risk of death in relation to the concentration of the interval-scaled variables ER and PgR was continually estimated by serial Cox regression analyses. Thus we circumvented loss of information due to primary categorization and avoided assumptions about relations between factor and risk. Based on 2,735 consecutively accrued primary breast cancer cases (median follow-up 56 months) we identified node-negative patients up to 60 years of age as the relevant subpopulation. High (>/=300 fmol/mg protein) ER concentrations exhibited an even more unfavorable impact (p < 0.03) on overall survival than ER concentrations of less than 10 fmol/mg protein. The well-known association of age and ER concentration was definitely excluded as an underlying biological cause for the increased risk. Differences in the distribution of other prognostic factors (HER-2/neu, Ki-67, DNA ploidy) were also excluded. As we observed a preponderance of pT2 tumors in the high ER group, we repeated the analysis, selectively focusing on pT2 tumors in the relevant subgroup, but the effect remained unchanged. In contrast, node-positive patients adjusted for age significantly (p = 0.02) profited from high ER concentrations as compared to the ER-negative group. As the phenomenon did not occur in node-positive patients, receptor defects in the high-ER group seem unlikely. To the contrary, we suspect that ER overexpressing cells are hypersensitive even to low levels of estrogens. Once they have sneaked past local barriers prior to primary surgery, they may cause early death in the absence of appropriate adjuvant endocrine therapy.