question about expression of two combined genes

Question

[rob.costa1234]

I am analyzing and RNAseq data and want to correlate expression of two genes (combined expression) in non-metastatic and metastatic tumors. My question is how do I divide expression of two genes can I take average of two genes in each sample and the assign a cut off to design them as high and low and then relate with clinical status? Is there another best approach instead of average? what is the common practice? Thanks