Assessment of existing bridge structures for inherent safety level or for lifetime extension purposes is often more challenging than designing new ones. With increasing magnitude and frequency of axle loads, reinforced concrete bridge decks are susceptible to fatigue failure for which they have not been initially designed. Fatigue verification and prediction of remaining service duration may turn out to be critical for civil infrastructure satisfying the required reliability. These structures are exposed to stochastic loading (e.g. vehicle loads, temperature loads); on the resistance side, reinforced concrete also behaves in a stochastic way. This paper presents a probabilistic reliability framework for assessment of future service duration, which includes probabilistic modelling of actions based on large monitoring data and probabilistic modelling of fatigue resistance based on test data. A case study for the steel - reinforced concrete slab of the Cret de l'Anneau Viaduct is presented along with calibration of resistance partial safety factors for lifetime extension.