Correcting for oversampling

What is the best method for correcting for oversampling?

I am looking at a simple exposure outcome relationship via linear regression. The cohort happens to be oversampled for one of the confounders of this (around 50% rather than ~5% as in general population from which the sample came).

How do I account for this when presenting my results? Should the sample be weighted? (And if so, how accurate does the general population measure need to be..)