Relationships between the binary dependent outcome variable of VTE with other patient-specific variables assessed by logistic regression analyses for all VTE (pseudo r² for final model 0.13, p<0.0001) and community-restricted VTE (pseudo r² for final model 0.12, p=0.0002). Stepwise regression based on the likelihood-ratio method was used to construct the models. Note that in each setting, factor VIII (FVIII) emerged as most significant in the first step, and no other variable was significant, once adjusted for FVIII levels in the 343 patients. In contrast to the FVIII regression analyses, in neither model was the series indicator variable significant, confirming the validity of pooling the series to enhance statistical power.