This study evaluated the ability of a chairside filtration system to remove particulate-based mercury (Hg) from dental-unit wastewater. Prototypes of the chairside filtration system were designed and fabricated using reusable filter chambers with disposable filter elements. The system was installed in five dental operatories utilizing filter elements with nominal pore sizes of 50 microm, 15 microm, 1 microm, 0.5 microm, or with no system installed (control). Daily chairside wastewater samples were collected on ten consecutive days from each room and brought to the laboratory for processing. After processing the wastewater samples, Hg concentrations were determined with cold vapor atomic absorption spectrometry (USEPA method 7470A). Filter systems were exchanged after ten samples were collected so that all five of the configurations were evaluated in each room (with assignment order balanced by a Latin Square). The numbers of surfaces of amalgam placed and removed per day were tracked in each room. In part two, new filter systems with the 0.5 microm filter elements were installed in the five dental operatories and vacuum levels at the high-velocity evacuation cannula tip were measured with a vacuum gauge. In part three of the study, the chairside filtration system utilizing 0.5 microm and 15 microm filter elements was evaluated utilizing the ISO 11143 testing protocol, a laboratory test of amalgam separator efficiency utilizing amalgam samples of known particle size distribution. Mean Hg per chair per day (no filter installed) was 1087.38 mg (SD = 993.92 mg). Mean Hg per chair per day for the 50 microm, 15 microm, 1 microm, 0.5 microm filter configurations was 79.13 mg (SD = 71.40 mg), 23.55 mg (SD = 23.25 mg), 17.68 mg (SD = 17.35 mg), and 4.25 mg (SD = 6.35 mg), respectively (n = 50 for all groups). Calculated removal efficiencies from the clinical samples were 92.7%, 97.8%, 98.4%, and 99.6%, respectively. ANCOVA on data from the four filter groups, with amalgam-surfaces-removed included as a significant covariate, was statistically significant (P < 0.0001). Tukey post-hoc comparisons (P < or = 0.05) indicated that the 50 microm filter removed less mercury than all other filters and the 0.5 microm removed more mercury than the 50 microm and 15 microm filters. Chairside vacuum measured on chairs with the 0.5 microm filters installed were minimally affected at the time of installation, and then gradually diminished as the filters became loaded with debris. The 0.5 microm configuration passed the ISO 11143 testing protocol at 96.8% efficiency.