Abstract The treatment of undesirable outputs in Data Envelopment Analysis (DEA) has received great research attention recently. As such and as are presented in this work, there are four possible options to deal with those: 1. ignoring them from the production function, 2. treating them as regular inputs, 3. treating them as normal outputs and 4. performing necessary transformations to take them into account. Also new model propositions for their treatment are being presented. Each method brings with it, benefits and drawbacks which each researcher should take into account at every stage of their research and assesses which method is more appropriate to be used.