We developed a quantitative real-time PCR assay targeting the mip gene of Legionella pneumophila for a prospective study from September 2004 to April 2005. It was compared with a standard culture method (French guideline AFNOR T90-431), analysing 120 water samples collected to monitor the risk related to Legionellae at Nantes hospital and to investigate a case of legionellosis acquired from hospital environment. Samples from six distinct water distribution systems were analysed by DNA extraction, amplification and detection with specific primers and FRET probes. The detection limit was 100 genomic units of L. pneumophila per liter (GU/l), the positivity threshold about 600 GU/l and the quantification limit 800 GU/l. PCR results were divided into three groups: negative (n=63), positive but non-quantifiable (n=22) or positive (n=35). PCR showed higher sensitivity than culture, whereas four culture-positive samples appeared negative by PCR (PCR inhibitor detected for two of them). Although no correlation was observed between both methods and real-time PCR cannot substitute for the reference method, it represents an interesting complement. Its sensitivity, reproducibility and rapidity appear particularly interesting in epidemic contexts in order to identify the source of contamination or to evaluate critical points of contamination in water distribution systems.