@LEXUS I think it would be benificial for you to read this paper written about CPA before you continue with taking high doses of the drug (read paragraphs from the right side to the left side starting with the introduction)
This information is important to keep in mind when you are taking CPA without doing research on it's phase trials. (im assuming you haven't read it before)
Title: The DNA damaging drug cyproterone acetate causes gene mutations and induces glutathione-S-transferase P in the liver of female Big BlueTM transgenic F344 rats.
Although the study was done on rats liver, it is still good information to have when you are considering very high doses of CPA. The term liver damage includes cancer, cell/gene mutation, high liver enzyme and the list goes on.
https://academic.oup.com/carcin/article-pdf/19/2/241/19258155/190241.pdf
See - "
Results Mutagenesis assay The experiments performed in this study reveal that CPA was a mutagenic compound in rat liver. In experiment 1, performed with one animal per dose and expression periods of 11 and 22 weeks, a dose of 100 mg CPA/kg b.w. was found to increase the mutation frequency determined in the controls by a factor of almost 4, while doses of 25 and 50 mg CPA/kg b.w. were ineffective, with the possible exception of 50 mg CPA/kg b.w.
11 week assay time (Table I). Experiment 2 performed with five animals per dose group and an expression period of 6 weeks revealed a dose dependent increase of mutation frequency at doses between 75 and 200 mg CPA/kg b.w. (Figure 1, Table I). Mutation frequencies determined at doses of 25 and 50 mg CPA/kg b.w. did not exceed those determined in control animals. For interpretation four common dose–response models were employed to fit the data.
For the 6 week assay, Figure 1 presents a plot showing the observed proportions of mutants of the individual rats, the pooled proportions and the four dose–response models fitted. The linear model (Model 1) produced the worst fit to the data. Models 2 and 4 are better than Model 1 and indistinguishable from each other. The linear E-NOEL model (Model 3) fits the data even slightly better than Models 2 and 4. The overall lack-of-fit statistics support these observations. For Model 1 the deviance has an associated p value of 0.03 indicating" (page 242)