Old bridges in the Netherlands are reassessed to prove their structural safety because of the increased traffic loads. The Eurocode model for the shear capacity of prestressed concrete beams with sufficient shear reinforcement was found to be very conservative for small amounts of transverse reinforcement, which is typical for old bridges. Code provisions based on the modified compression field theory appear to be much more accurate (Bentz, Vecchio, & Collins,2006). On the other hand, these models make no distinction between flexural shear and shear tension failure, while 25% of the bridges consists of T-, I-, or box beams which are sensitive to shear tension failure. The aim of this thesis is to present a more accurate prediction for shear tension failure based on the modified compression field theory (MCFT). The report focuses on both the CSA-model and Response-2000. The CSA-model is a simplification of the MCFT for beams under the assumption that σ_z=0. The CSA-model shear resistance consists of a concrete and steel part. Response-2000 is a cross-sectional analysis program that works as a non-linear finite element analysis and not only takes into account bending but also shear.Response analysis of 32 beams (experiments of Xie (Xie, 2009), Choulli (Choulli, 2005), Hanson (Hanson & Hulbos, 1965) and Leonhardt (Leonhardt, Koch, & Rostasy, 1973)) have been made and compared with the experimental observations. The failure load is predicted with a mean ratio of 1.38 and a COV of 19% compared to the experiments. It was found that for small a/d-ratios predictions are more conservative. Failure mechanisms: rupture of the stirrups, crushing of the web concrete and slipping/major crack opening are found in Response. 87% of the failure mechanisms was predicted correctly compared to the experiments. The critical cross-section compared with the experimentally observed failure zone showed that 60% of the beams was predicted in the failure zone. The shear force in the cross section is resisted by reinforcement steel, aggregate interlock and the uncracked concrete, which were found to take up respectively 1/2, 1/6 and 1/3 of the shear force. The steel part is predicted accurately while the aggregate interlock part is underpredicted due to an over prediction of the crack spacing. The uncracked part is overpredicted for small amounts of reinforcement and underpredicted for larger amounts of reinforcement. The data of the Response analysis have been used to modify the CSA-model to a model solely describing shear tension failure. The comparison showed that the CSA-model underestimates the steel part, overestimates the concrete part, takes a to large cracked height and doesn't take into account the contribution of the uncracked part. From this comparison, modifications for β,θ,V_max,h_crack and ϵ_x are proposed using Response data and proposals from Esfandiari (Esfandiari & Adebar, 2009). The flanges are taken as uncracked and the contribution of the uncracked part is described with a linear elastic shear stress distribution. This has led to a proposal for a model that solely describes shear tension failure, where the most important addition is the contribution of the uncracked height. Calculations show that the model gives conservative predictions compared to the experiments (mean ratio of 1.36 and a COV 22%) and the predictions are almost the same as for Response. The parameters are predicted conservative compared to the Response-2000 predictions.