Associations between type of product and inhibitors in previously untreated patients (PUPs) with severe hemophilia: switches and particular products can disturb analysis

To the editor:

Genetic and environmental risk factors for inhibitor development in severe hemophilia A have been reported in various observational studies. The CANAL (Concerted Action on Neutralizing Antibodies in severe hemophilia A) historical cohort studied the effect of product type by searching for differences between 1) recombinant factor VIII (rFVIII) and plasma-derived factor VIII (pdFVIII) and 2) depending on the concentration of von Willebrand factor (VWF).¹  Despite a large number of patients (316 previously untreated patients [PUPs]) and a multivariate survival analysis, no difference was found. Regarding the first objective, Gouw et al observed an adjusted relative risk (aRR) rFVIII/pdFVIII of 1.4 (confidence interval [CI], 0.9 to 2.5)¹  while we found an aRR of 2.4 (CI, 1.0 to 5.8),²  and a recent English study reported an adjusted odds ratio of 1.83 (CI, 0.9 to 3.72).³  The confidence intervals overlap, but 2 differences could account for the RR gradient observed. First of all, Gouw et al considered product type as a time-dependent covariate. The definition of switches was not clearly specified, but at least 54 patients, including 49 (36%) of the patients initially treated with pdFVIII, changed product early (after a median of 5 cumulative exposure days [CEDs]), probably for rFVIII in most cases.¹  These patients were kept in the analysis until 50 CEDs and were included in the estimation of inhibitor incidence with rFVIII. This analysis option is based on the hypothesis that product-associated risk is only determined by the product received at the last injection. However, according to Dasgupta et al, VWF protects FVIII from endocytosis by human dendritic cells and subsequent presentation to FVIII-specific T cells.⁴  If VWF provides protection during the initiation of treatment, this analysis option underestimates the rFVIII/pdFVIII RR and, by the same mechanism, could hinder study of the effect of VWF concentration. Chalmers et al took into account the appearance of inhibitors within the first 50 CEDs and studied the effects of the initial FVIII treatment.³  If patients changed product before 50 CEDs (which is probable), the difference between the treatments could also have been underestimated. Using a nonexperimental approach, we think that the best strategy is to consider product type as a fixed cofactor and to not take into account follow-up after the first switch. Secondly, in the CANAL study, the patients received 23 different pdFVIIIs, including 1201 CEDs to Beriate,¹  which represents 38% of the CEDs to pdFVIII “containing considerable quantities of VWF.”¹  However, this product has very low VWF concentration (0.09 IU/IU FVIII),⁵  and it appears that the immunoprotective effect of VWF is concentration dependent.⁴  Furthermore, in FVIII knockout mice, this product was as immunogenic as 2 first-generation rFVIII products.⁵  Thus, this pdFVIII could be associated with a particular immunogenicity, and it would be interesting to perform a sensitivity analysis excluding patients having received it in the rFVIII/pdFVIII comparison and to include testing of the classification of Beriate with pdFVIII products with low VWF content to study the effect of VWF concentration. The possibility that certain pdFVIII products could be less immunogenic and, most importantly, identification of the physiopathological mechanisms of these possible differences remain major issues for the development of new FVIIIs.