Microstrip filters are widely used in modern microwave communication systems such as Wi-Fi, 5G and satellite navigation. This increasingly broad range of applications implies that the frequency spectrum becomes more and more crowded, resulting in more demanding filter specifications. To meet these design specifications, computationally expensive electromagnetic simulations are often used.To speed up the filter design flow, we introduced metamodels for design optimization and variability analysis tasks. This paper illustrates the advantages of using metamodels in filter design by use of a defected ground structure. The average CPU time of the design process decreases drastically, while achieving accurate results.