Twenty case studies, involving the application of fracture mechanics to structural integrity, have been reviewed and compared to a similar report published in 1978. Sixteen of the new cases discussed failures, while four were fitness-for-purpose analyses (i.e. evaluation of safe operating conditions of defect-containing structures). In reviewing the case studies, it was found that the calculated value of stress intensity at failure was usually only approximately equal to the reported value of fracture toughness. Furthermore, in a number of cases, the calculated stress intensity was significantly less than the reported fracture toughness, thereby indicating a non-conservative fracture mechanics analysis. The probable cause for this relatively poor correlation was that the inputs into the analyses, particularly fracture toughness, were often appro~mations. Both reviews suggest that the likelih~ of failure is particularly large when there is a defect greater than 25 mm (1 .O in.) in size and when the fracture toughness: yield strength ratio is less than 0. I6 ,/m (1 .O Jin.). No significant improvement in accuracy of failure analysis was detected compared to the earlier paper.