Matrix metalloproteinases (MMPs) play an important role in tissue remodelling associated with various physiological and pathological processes, such as morphogenesis, angiogenesis, tissue repair, arthritis, chronic heart failure, chronic obstructive pulmonary disease, chronic inflammation and cancer metastasis. As a result, MMPs are considered to be viable drug targets in the therapy of these diseases. Despite the high therapeutic potential of MMP inhibitors (MMPIs), all clinical trials have failed to date, except for doxycycline for periodontal disease. This can be attributed to (i) poor selectivity of the MMPIs, (ii) poor target validation for the targeted therapy and (iii) poorly defined predictive preclinical animal models for safety and efficacy. Lessons from previous failures, such as recent discoveries of oxidative/nitrosative activation and phosphorylation of MMPs, as well as novel non-matrix related intra- and extracellular targets of MMP, give new hope for MMPI development for both chronic and acute diseases. In this article we critically review the major structural determinants of the selectivity and the milestones of past design efforts of MMPIs where 2-/3-dimensional structure-based methods were intensively applied. We also analyse the in vitro screening and preclinical/clinical pharmacology approaches, with particular emphasis on drawing conclusions on how to overcome efficacy and safety problems through better target validation and design of preclinical studies.