The first thing for you to do is prove that whatever expression you have is jointly concave *or jointly convex) in all the variables. It matters very much which are variables and which are constants (input data). And then you need to use that expression in a manner which corresponds to which of concave or convex it is.
Of course you already know this because you read very carefully the link in my preceding post. Or if you did not, you will now.
If your expression (and overall optimization problem) does not meet these needs, CVX is nto the right tool for your problem, in which case you could consider using YALMIP with a non-convex solver.