Even if fouling in membrane bioreactors (MBRs) has been extensively studied during the last decade, its causes and mechanisms are not well understood yet. Furthermore, few full-scale and long-term experiments have been published, and their results do not always match with the models developed from lab-scale studies. A statistical approach linking long-term and short-term permeability evolution with operational variables in full-scale membrane bioreactors for domestic waste-water treatment is presented. Data originate from a 66,700 P.E. MBR plant monitored for more than one year. Permeability and several fouling indicators were calculated in each of the four hollow-fibre membrane tanks of the plant. The influence of SRT, temperature, MLSS, F:M ratio, iron dosing and membrane flux on daily permeability evolutions, instantaneous permeability evolution and hydraulic backwash efficiency was studied. In order to minimise the bias due to correlations between input variables, a statistical approach using principal component regression and partial least-square regression was tested. Flux, temperature, SRT and F:M ratio are the most influential input variables on long-term permeability evolutions. Iron dose and MLSS are less correlated with fouling indicators. The proposed approach may be improved by integrating the history of the membrane to better describe and predict the permeability evolution.