In this paper link adaptation for interference alignment (IA) based on imperfect channel state information (CSI) is investigated. We consider a MIMO interference channel where the transmit and receive spaces are determined by IA. We look at maximizing a weighted sum of the average rates provided that a certain set of bit-error-rate and power constraints are satisfied by dynamically adjusting coding, modulation and powers. The problem is quite general and intractable. We resort to some approximations and provide simulations to show the accuracy of the approximations in the regimes with practical interest.