In drinking water supply, arsenic exists mostly as As(III) and As(V) forms which are toxic to living organisms. According to WHO recommendations, the drinking water standard was reduced from 50 to 10 μg.L-1 and many regulatory agencies have recently accepted this new standard. Most of the existing treatment processes are effective only on As(V) which forms anionic complexes unlike As(III) which forms mobile neutral complexes. As a consequence, the removal of arsenic from water requires a preliminary oxidation step of As(III) form which can be done by using Bayoxide, GEH but their cost remains high for small drinking water treatment units. This could be overcome thanks to biological treatment using As(III)-oxidising bacteria associated with As(V) adsorption process. Numerous autotrophic bacteria are able to oxidise arsenic. Among them, Thiomonas arsenivorans is able to oxidise As(III) up to 100 mg As(III).L-1 and appears to be an interesting bacterium in that it is an autotrophic bacterium which uses arsenic as an energy source and carbon dioxide or carbonates as carbon source The objectives of this work were (1) to optimise the As(III) oxidation from water drinking resources in fixed bed bioreactor and (2) to study the biofilm development during the treatment in order to improve the efficiency of the treatment. Experiments were performed in plastic bioreactors equipped with intermediate water outflows, filled with pozzolana sieved at 5-7 mm and inoculated with a culture of T. arsenivorans. Bioreactors were then continuously up-flow fed with As(III) and a synthetic water at pH 7 whose composition and temperature (12°C) are closed to that of a natural groundwater. Residence time (30 and 15 minutes) and As(III) concentrations (15 and 50 μg.l-1) were studied. Arsenic speciation and analysis was performed using GF-AAS. At the end of the experiment, bioreactors were dismantled for biolfim analysis: bacterial counting, bacterial cell viability, bacterial diversity by CE-SSCP, diversity of aoxB gene (As(III)-oxidase enzyme) by PCR-DGGE were performed on the pozzolana samples.