The results of Reactivity-Initiated Accidents (RIA) experiments have been analysed and the main variables affecting the fuel failure propensity identified. Fuel burn-up aggravates the mechanical loading of the cladding, while corrosion, or better the hydrogen absorbed in the cladding as a consequence of corrosion, may under some conditions make the cladding brittle and more susceptible to failure. Experiments point out that corrosion impairs the fuel resistance for RIA transient occurring at cold conditions, whereas there is no evidence of important embrittlement effects at hot conditions, unless the cladding was degraded by oxide spalling. A fuel failure threshold correlation has been derived and compared with experimental data relevant for BWR and PWR fuel. The correlation can be applied to both cold and hot RIA transients, account taken for the lower ductility at cold conditions and for the different initial enthalpy. It can also be used for non-zero power transients, provided that a term accounting for the start-up power is incorporated. The proposed threshold is easy to use and reproduces the results obtained in the CABRI and NSRR tests in a rather satisfactory manner. The behaviour of advanced PWR alloys and of MOX fuel is discussed in light of the correlation predictions. Finally, a probabilistic approach has been developed in order to account for the small scatter of the failure predictions. This approach completes the RIA failure assessment in that after determining a best estimate failure threshold, a failure probability is inferred based on the spreading of data around the calculated best estimate value.