Is a two-way anova enough?

Hello everyone.
As I am far away from being an expert in biostatistics and get confused by it easily (I will do some training in the next year and hopefully get better at it), I would like to ask for advice.

The study setup is the following:

Data all is gene expression data generated by qPCR

We wanted to elude the effects of a compound on the carcinogenic process in rats, so there are two major groups: rats that recived the compound, and rats that recieve a control diet.
To induce carcinogenesis, rats recieved a hormone. We recieved tumors as well as tissue that was not a tumor yet despite recieving the hormone (of which we do not know if it is really normal, or in a pre malignent state).

So we have 2 major groups, with 3 minor groups in each groups (number of samples in brackets):

A) Control diet
  1. Normal tissue, no hormone treatment (19)
  2. Normal or pre malignet tissue, with hormone treatment (10)
  3. Tumor tissue, with hormone treatment (16)

B) Compound diet
  • Normal tissue, no hormone treatment (9)
  • Normal or pre malignet tissue, with hormone treatment (8)
  • Tumor tissue, with hormone treatment (7)

So far, we are using 2-way-Anova grouped analysis in Prism 5 for statistical analysis, but we are not convinced that this is the right way to go since we want to compare a lot of things here:
1. Effect of the diet (with or without compound)
2. Effect of carcinogenesis (normal or Tumor)
3. Effect of the hormone (normal/pre malignent that recieved the hormone and tumors)

We are comaring 3 things here, so we should use 3-way anova, but 2 of the variables are somehow connected (for the hormone treatment we have 2 groups).
I read that for 3-way anova you need to have 3 independant variables.

I would be very gratefull for advice.
If I forgot to mention something, please let me know.

Best wishes and thank you very much!