While androgen and androgen receptor (AR) activity have been strongly implicated in prostate cancer development and therapy, the influence of the CAG repeat, which is found within the first exon of the AR gene, on prostate carcinogenesis is still unclear. We investigated the differences in the length of the CAG repeat between prostate cancer patients and controls in the Chinese population as well as between TMPRSS2:ERG fusion positive and negative samples. A general association between prostate cancer and either longer or shorter AR CAG repeat length was not observed in the Chinese population. However, our data suggest that certain CAG repeat lengths may increase or decrease prostate cancer risk. Shorter CAG repeat length was also not shown to be associated with a higher induction rate of TMPRSS2 and ERG proximity, an essential step for TMPRSS2:ERG fusion formation. However, samples with a CAG repeat of 17 were found more frequently in the TMPRSS2:ERG fusion positive than negative prostate cancer cases and mediated a higher rate of androgen-induced TMPRSS2 and ERG co-localisation than AR with longer (24) and shorter (15) CAG repeats. This suggests that 17 CAG repeats may be associated with TMPRSS2:ERG fusion positive prostate cancer, but may have a preventive role for prostate cancer in the Chinese population, which has a low TMPRSS2:ERG fusion frequency. This study suggests that different mechanisms for the association of CAG repeat length polymorphism and prostate cancer exist in different ethnic populations.