Data regarding the efficacy of programmes to control meticillin-resistant Staphylococcus aureus (MRSA) in intensive care units (ICUs) are limited. We performed an observational 'before-and-after' study to evaluate the search-and-destroy (S&D) strategy as compared with S&D and isolation (SDI), to control MRSA in a general ICU. S&D included active surveillance, contact precautions and treatment of carriers; in SDI, isolation or cohorting were added. Three phases were identified: period 1 (p1), 1996-1997, before the introduction of programme; period 2 (p2), 1998-2002, with S&D programme; period 3 (p3), 2003-2005, with SDI in a new ICU. During the 10 years of the study we observed 3978 patients; 667, 1995 and 1316 patients in p1, p2 and p3 respectively. The numbers of MRSA-infected patients were 19 in p1, 23 in p2, and 6 in p3. The infection rate was 3.5, 1.7 and 0.7 cases per 1000 patient-days in p1, p2 and p3, respectively; a significant reduction was observed between p1 vs p2 (P=0.024) and p2 vs p3 (P=0.048), although the latter was not confirmed by a segmented regression analysis. The proportion of ICU-acquired MRSA cases was 80%, 77% and 52% during p1, p2 and p3, respectively (P=0.0001 for trend). The proportion of S. aureus isolates resistant to meticillin was 51%, 32% and 23% during p1, p2 and p3, respectively (P<0.0001 for trend). S&D strategy was effective in significantly reducing MRSA infection, transmission rates and proportion of meticillin resistance in an ICU with endemic MRSA. SDI may further enhance S&D efficacy.