The exon 1 of the human androgen receptor gene (AR) contains both CAG (polyglutamine) and GGN (polyglycine) repeat length polymorphisms. Large CAG repeats have been related to an increased risk of breast cancer (BC), whereas the influence of the GGN repeats is still unclear. Here, we have studied how the length of CAG and GGN repeats is associated with the risk of BC in a population from Tenerife (Canary Islands, Spain). The study was carried out on 257 woman diagnosed with BC and 393 controls, nesting in the 'CDC of the Canary Islands' cohort study. The AR CAG and GGN genotyping was performed by means of PCR amplification with specific fluorescently labelled primers followed by a capillary electrophoresis. The allelic distribution of CAG and GGN polymorphisms was similar in cases and controls. The mean of short and long CAG and GGN alleles did not show differences between cases and controls and the same was true when the average length of both CAG alleles (CAG(n)) and GGN alleles (GGN(n)) was considered. However, when CAG(n) and GGN(n) were categorised using 22 and 24 repeats as the cut-off point, respectively, significant differences between cases and controls were observed. The CAG(n)>22 repeats were more frequent in cases than in controls, being associated with an increased risk of BC (OR=1.49; CI(95%)=1.06-2.09; p=0.021). No significant differences were found for categorised GGN(n). For CAG(n)/GGN(n) combinations, the highest BC risk was found to be associated with the CAG(n)>22/GGN(n)24 combination (OR=2.47; CI(95%)=1.37-4.46; p=0.003). In conclusion, our results indicate that longer CAG(n)/GGN(n) combinations increase the risk of BC and suggest that CAG and GGN AR polymorphisms should be considered in order to assess the BC risk.