Statistical matching aims at combining data available in distinct sample surveys referring to the same population. This objective can be achieved at micro level, ie an integrated data set is produced, or at macro level, ie a parameter is estimated. When data come from complex sample surveys, the application of statistical matching techniques poses different problems. Existing approaches make use of weight calibration methods, others are based on the concatenation of the two data sources, others more consist in using weights only at the inferential stage while the matching procedure is performed as if the sampling design would be a simple random sampling. In this paper an extension of the latter methods considering the weights also in the matching phase is investigated. A comparison of different techniques focused on micro level statistical matching is carried out by using simulated data.