Due to their aesthetic and structural advantages, tubular space truss structures are enjoying increasing popularity in modern bridge construction. The use of cast steel nodes for the joints between the circular hollow section members is also becoming increasingly popular. The fatigue design of such joints however requires additional knowledge with respect to their fatigue resistance. The present work deals with the global fatigue behaviour of cast steel nodes used in longitudinal truss girders of steel-concrete composite bridges. The global fatigue behaviour of cast steel nodes in a truss girder is quantified on the basis of experimental investigations. The relative influence of the resistance of the cast steel node and the resistance of the welds is analysed as a function of various parameters. An economically optimal fatigue design consists of adapting the fatigue resistance of the cast node to that of the welds. The experimental results show very clearly that the fatigue behaviour is governed by the welds in all tested configurations. Their fatigue behaviour is therefore investigated in a research program conducted simultaneously. The present work concentrates on the adaptation of the resistance of the cast steel nodes to that of the welds. Consequently, allowable initial casting defect sizes must be defined as a function of the required fatigue resistance of the welds. Using a numerical boundary element model, they are calculated for cast steel nodes in a typical steel-concrete composite bridge. The results show that brittle failure of a node containing cracks does not occur. A crack depth equal to 90 % of the wall thickness at the crack location is therefore chosen as the failure criterion, assuming that a through-thickness crack is unacceptable. Varying between 28 and 88 % of the wall thickness, the resulting allowable initial casting defect sizes are very large. Aiming at a general application for steel-concrete composite bridges, the results of the numerical investigations are represented by an approximate formula for the stress intensity factor based on a constant correction factor. This approximation results in a considerable simplification of the procedure for the fatigue design of cast steel nodes established for the aforementioned typical composite bridge. The approximate formula is used to perform a parametric study. The influences of the utilisation ratio under traffic and fatigue loads, the cast steel fracture toughness and yield strength and the node dimensions on the defect size are described. Assuming a fracture toughness likely to be encountered in practice and a mean utilisation ratio of the node as well as a range of node dimensions, the allowable initial casting defect sizes are quantified. The present work shows that the fatigue resistance of the welded joints needs to be improved substantially in order to benefit from the high fatigue resistance of the cast nodes. Weld details including a backing ring offer a higher fatigue resistance than those without. Using the proposed design concept, allowable initial defect sizes in cast steel nodes can be estimated as a function of the required fatigue resistance.