The clinical phenotype of patients with haemophilia A (HA) often differs between individuals with the same factor VIII (FVIII) gene defect (e.g. within the same family) or the same coagulant activity of FVIII (FVIII:C). We proposed that because the thrombin generation assay in platelet-poor plasma of HA patients provides more information [peak thrombin concentration, endogenous thrombin potential (ETP), rate of thrombin generation and lag-time] than a clot-based FVIII assay it might provide insight into these differences. We therefore investigated the relation between the results of the thrombin generation assay and the clinical severity in nine families with HA (23 patients with different phenotypes). We also examined the contribution of prothrombotic risk factors: (FV Leiden G1691A and prothrombin G20210A), the coagulant activity of FVIII and tissue factor (5'UTR) polymorphisms. Our data detect marked differences between individuals but these did not correlate with the reported clinical phenotype. These differences were also reflected in a marked difference in response to the therapeutic amounts of FVIII. This might account for differences in amounts of treatment consumption. Reduced peak and possibly rate of thrombin generation, rather than FVIII:C or ETP appear to represent the critical defects in FVIII-deficient plasma. We suggest that the analysis of parameters in thrombin generation is a useful tool to detect bleeding tendency in HA but not to predict the modulation of the haemorrhagic tendency in patients within families. However the presence of the other factors such as vessel wall components, protein C and platelets might need to be incorporated into this system.