Several case-control studies have examined the possibility that minor DNA-sequence polymorphic variants of p53 are human cancer risk factors. Generally, such common polymorphic DNA-sequence variants, which can attain a frequency of around 10%, are considered to be benign. However, in our recent preliminary report we found a specific p53 allele to be associated with modest breast cancer risk among Caucasian women. This report, expanded to include more women and using a new analytic approach, contains data that suggest a strong gene-dose effect of the p53 variant on breast cancer risk. P53 haplotypes defined by 3 polymorphisms (codon 72, intron 3, and intron 6) have been implicated in breast cancer]. Our recent studies used a novel, practical method to determine true extended haplotypes of these three polymorphisms in diploid genomes. Interestingly, we identified a haplotype composed of the three minor variants as the third most common of 8 possible, with two haplotypes absent (analysis of >750 chromosomes). This haplotype, designated 1-2-1 after the original nomenclatlure, was found to be overrepresented among Caucasian breast cancer patients compared to age- and race-matched control patients, particularly in postmenopausal women (OR = 2.5, 95 % C I = 1.3 -4.8; p = 0.009; n = 52 case patients and 88 control patients). We have now examined the effect of gene-dosage, in a larger group but including the same Caucasian women, using an analysis for linear trend in proportions. Among postmenopausal women, an increased risk with gene-dosage was observed (OR-heterozygotes = 2.7, OR-homozygotes = 4.2; age-adjusted X2 for trend = 6.05, P = 0.01). The data suggest that the penetrance of the 1-2-1 allele is low. However, if its frequency in our hospital based control population is representative of the general population, inheritance of this allele is likely to carry a substantial attributable risk. The haplotype frequencies in our control series (0.10) are similar to estimates made by Sjalander et al. in a population-based study of Swedish subjects (0.13), suggesting that this is the case. Gene-dose effects have previously been noted for HRAS-l rare alleles and polymorphisms in carcinogen metabolizing enzymes. As far as we are aware this is the first report of a gene-dose effect for a polymorphism in a tumor suppressor gene. Similar effects were not detected in African-Americans or Hispanics, although the frequency of the 1-2-1 haplotype in the control patients was found to be higher 0.28 and 0.19, respectively than in Caucasian control patients (0.11). However, the numbers of study subjects were small (22 African-American patients and 37 Control patients; 28 Hispanic patients and 62 control patients). These data suggest that other endogenous or exogenous factors, perhaps related to race, contribute to breast cancer etiology in combination with this p53 haplotype.