In the present report, the design of a new dialysate delivery system to produce low to moderately contaminated dialysate is described. In addition, the first data on bacterial counts and lipopolysaccharide (LPS) concentrations in both the dialysate and the blood during hemodialysis (HD) with superflux dialyzers are presented. In this prospective study, 37 patients were randomized into two consecutive periods of 12 weeks to HD with a high flux polysulfon (PS), a superflux PS, a superflux cellulosic tri-acetate (CTA) or a superflux CTA dialyzer with filtered dialysate (CTAf), resulting in 74 periods in which measurements were obtained. Filtered dialysate showed significantly lower bacterial counts, if compared with nonfiltered dialysate (p < 0.001). As for LPS, marked differences were not observed between filtered and nonfiltered dialysate, whereas mean plasma LPS concentrations were below the value of the dialysate at all time points (p < 0.001). Plasma LPS concentrations decreased significantly during HD with all four modalities (F 60: t0 0.032 ± 0.005, t180 0.026 ± 0.009 endotoxin units (EU)/ml, p = 0.001; F 500S, t0 0.031 ± 0.004, t180 0.027 ± 0.005 EU/ml, p = 0.001; Tricea 150G: t0 0.032 ± 0.004, t180 0.025 ± 0.005 EU/ml, p < 0.001; and Tricea 150Gf: t0 0.034 ± 0.007, t180 0.025 ± 0.006 EU/ml, p < 0.001). During HD with highly permeable dialyzers and moderately contaminated dialysate, plasma LPS concentrations decreased significantly, irrespective of the material used (PS or CTA), the flux characteristics of the devices (high flux or superflux), or the presence of a bacterial filter.