The quality of the RNA extraction system plays a crucial role for the detection of viruses in water or environmental samples. In the present study we investigated the detection limit, the efficiency and the presence of eventually co-extracted inhibitors by comparing four commercially available large scale (>or=1 ml) viral RNA extraction methods (QIAamp Viral RNA Mini Kit in combination with preconcentration by Centricon YM-100 [Centricon-QIAamp], QIAamp UltraSens Virus Kit, NucliSens Isolation Kit and NucleoSpin RNA Virus F). A 1 ml 50 mM glycine (pH 8.0) containing 1% beef extract was spiked with different concentrations of poliovirus vaccine strains, extracted by the four methods and analysed by RT-nested PCR or RT-quantitative LightCycler PCR. Eight replicates were analysed for each concentration on different days. The positive cut-off point was determined to be at 0.25 CCID(50) per ml (Centricon-QIAamp), 1.46 CCID(50) per ml (UltraSens), 0.4 CCID(50) per ml (NucliSens) and 3.03 CCID(50) per ml (NucleoSpin). Quantitative analysis (LightCycler) of a high-titer sample showed significant differences between the efficiencies of the four extraction methods examined. The efficiencies of the extraction methods were normalized to the NucliSens method as follows: (71% Centricon-QIAamp, 18% UltraSens, 100% NucliSens and 23% NucleoSpin). In addition, spiked negative controls did show significant differences, indicating a co-extraction of inhibitors. Compared with the non-inhibited positive control the inhibitions were 21, 37, 27 and 68% for the Centricon-QIAamp, UltraSens, NucliSens and NucleoSpin methods, respectively. Taken together, these findings indicate that of the four evaluated extraction methods both the NucliSens and Centricon-QIAamp are best suited to extract viral RNA from water samples previously concentrated and have shown to be very sensitive, efficient and robust methods.