Deep-water tendon and riser systems are often subjected to severe fatigue loading from waves, currents and vessel movements. The girth welds between successive lengths of pipe or at pipe terminations are the locations most vulnerable to fatigue damage and accurate and reliable assessment of the fatigue performance of these welds is of significant importance. These welds are normally designed on the basis of appropriate S-N curves. In addition, it is common practice to perform an Engineering Critical Assessment (ECA) based on fracture mechanics fatigue crack growth principles, principally in order to establish acceptance limits for weld flaws that are consistent with the fracture toughness of the welds and the required fatigue life. However, for the most important case of fatigue failure from the weld toe, the mode of failure principally covered by the S-N curves, a conventional ECA invariably results in tolerable weld toe flaws that are too small for reliable detection by currently available NDE methods. In fact, this situation is not unreasonable since it is known from careful metallurgical examinations that weld toe fatigue cracks initiate at tiny (<0,5 mm deep) sharp imperfections that are an inevitable consequence of welding, whereas the detection limit for even the most sensitive NDE techniques is around 1–2 mm. Hence this paper discusses different models for including crack initiation and early growth of defects in girths weld, in addition some trends found in full-scale data from an on-going girth weld fatigue JIP is shown.