If you believe there are 'extraneous' effects / unwanted sources of variation in the dataset that could bias your results, then you could perform Surrogate Variable Analysis, which will essentially identify a set of variabes that 'capture' this unwanted variation. The idea, then, is to include these variables in your design forumla for the purpose of controlling for these.

Kevin