Recombinant factor VIII products and inhibitor development in previously untreated boys with severe hemophilia A

Hemophilia A (HA) is a hereditary coagulation disorder resulting from factor VIII (FVIII) deficiency.¹  Treatment consists of infusions of FVIII concentrates prepared from human plasma or by genetic engineering.²  Some patients develop neutralizing antibodies against these products, mostly within the first 50 exposure days (EDs). These so-called inhibitors may jeopardize the patient’s life³  and make therapeutic management more complex^4,5  and costly.⁶  Inhibitors arise in 15% to 35% of children with severe HA.^7,8  Several genetic and nongenetic risk factors for inhibitor development have been described. The main nongenetic risk factors are related to the modalities and circumstances of replacement therapy, such as age at treatment initiation, the FVIII product used, treatment intensity, prophylaxis regimen, major bleeds, and surgical procedures.^9-11  Research into inhibitor development has focused on the FVIII source—recombinant (r) vs plasma-derived (pd) products—with mixed results.^12,13  In 2004, a cohort study was launched in 29 hemophilia treatment centers (HTCs) in Europe, Israel, and Canada by the Research of Determinants of Inhibitor Development (RODIN) study group, to identify risk factors for inhibitor development in children with severe HA born between 2000 and 2009. The first article describing associations between FVIII products and inhibitor development in 574 consecutive patients was published in January 2013.¹⁴  The only significant result was an unexpectedly higher risk of inhibitor development with a second-generation full-length rFVIII product (Product D) compared with the most widely used third-generation rFVIII product. This result perplexed prescribers and hemophilia patients worldwide.^15,16  The European Medicines Agency quickly launched a review of the RODIN results.¹⁷  In December 2013, European Medicines Agency concluded that the available data did not support a higher inhibitor incidence with Product D than with other products. However, no new data challenged the RODIN findings, and it thus remained to be shown whether the observed difference was real or simply due to bias or chance (sampling fluctuations). In France, a prospective cohort was created in 1994 by the public health authorities to monitor hemophilia treatment safety.¹⁸  A subcohort of previously untreated patients (PUPs) was established to investigate risk factors for inhibitor development. Here we compared inhibitor incidence rates across rFVIII products in these PUPs with severe HA.

Almost all hemophilia patients in France have been included in the Réseau FranceCoag (hereafter, FranceCoag) national pharmacosurveillance network, which is based on voluntary participation of all French HTCs without support from pharmaceutical companies. Children included in the PUP cohort are closely monitored until age 18 years. Children born before 2000 were eligible if they had fewer than 4 EDs to FVIII, whereas those born from 2000 onward are included at diagnosis or as early as possible, provided their first infusions are traceable. The research into the differences between inhibitor development incidences in rFVIII and pdFVIII products is still relevant. However, the RODIN findings prioritized a comparison of rFVIII product immunogenicities to determine whether or not this class is homogeneous. We therefore selected boys with FVIII activity below 0.01 IU/mL enrolled in the FranceCoag PUP cohort who were first treated with rFVIII products. To facilitate the interpretation of the results of both studies, patients who also participated in the RODIN study were excluded from our analyses. The protocol for our observational cohort was approved by the French data protection watchdog. Parents or legal guardians received detailed written information about the objectives and modalities of the follow-up and were asked to approve their child’s enrollment in accordance with the Declaration of Helsinki.

Data were collected on electronic case report forms. Baseline data comprised demographic characteristics, HA-related medical history from birth, the F8 gene defect, family history of hemophilia and inhibitors, and ethnic origin. At each follow-up visit, major events having occurred since the last visit were collected (severe bleeds, hemarthrosis, surgical procedures, vaccinations) along with detailed hemophilia treatments. The results of FVIII and inhibitor assays performed during follow-up were also recorded. The collected data were automatically checked and repeatedly monitored for completeness and inconsistencies by 3 dedicated clinical research assistants in the HTCs. Independently from the FranceCoag Network, the French Hemophilia Association publishes a customized booklet for collecting the adhesive labels of all injected FVIII or FIX products to obtain details of home treatment. For several decades, this tool has been distributed to hemophilia patients in all French HTCs. In addition to data collected during hospitalization, this booklet allowed us to record the first 75 EDs in the same way as in the RODIN study, including the infusion dates, FVIII product brands and doses, indications, bleeding events, and types of surgery. These additional data were collected by a fourth clinical research assistant, independently of the basic data collection. Thus, any discrepancy between the 2 recording systems (regarding the date of the first infusion, FVIII products received, number of EDs at the date of inhibitor detection, initiation of regular prophylaxis, surgical procedures, and severe bleeding episodes) triggered extensive investigations based on the original files.

Patients enrolled in the FranceCoag Network are followed indefinitely, but only the first 75 EDs were considered for this analysis. The cutoff date was May 19, 2014. For patients who developed an inhibitor, follow-up ended on the last ED before its detection. The follow-up was censored after the last rFVIII infusion if the patient switched to a pdFVIII before ED75 or if the patient had not reached 75 EDs at the last clinical visit. The inhibitor assays were performed in the laboratory of each HTC. The primary outcome measure was inhibitor development during the first 75 EDs, defined as a positive assay result (>0.6 Bethesda units (BU)/mL) on any 2 dates. A secondary outcome measure was inhibitor development during the first 75 EDs with a peak titer of at least 5 BU/mL at any time.¹⁹  A third outcome measure was added to reflect the therapeutic impact of inhibitors, namely inhibitor development during the first 75 EDs treated at any time with a bypassing agent and/or an immune tolerance induction (ITI). All clinically significant cases were validated by an ad hoc clinical committee using a standardized procedure (supplemental Methods, available on the Blood Web site).

The main risk factor that was investigated was the rFVIII product received during the first 75 EDs. Six rFVIII products (A to F) have been marketed worldwide (Table 1). Products A to C were withdrawn from the European Union market and were replaced with new-generation rFVIII products (D to F) (supplemental Figure 1). Because patients could be switched from one product to another during the first 75 EDs, the rFVIII product was analyzed as a time-varying factor. As in the RODIN study, Product E served as reference.

We constructed a Cox proportional hazards model with ED as the observational time unit. The best-acknowledged fixed risk factors—the F8 gene defect, family history of hemophilia and inhibitors, ethnic origin, and age at first rFVIII infusion—were systematically included in the multivariate models, regardless of their statistical association with inhibitor development in this analysis. The following time-varying risk factors were also considered: calendar period, regular prophylaxis, treatment intensity markers as used in the RODIN study (interval between EDs and rFVIII dose calculated over the last 5 EDs, and peak treatment episodes), history of surgery,¹⁴  and history of severe bleeding. Time-varying risk factors associated with the all inhibitors outcome with a P value <.2 were retained in the final multivariate models. Crude and adjusted hazard ratios (aHRs) are reported for the association between a given rFVIII product and clinically significant inhibitor development. Associations were considered significant if P < .05. Cofactor definitions, grouping, and missing data procedures are described in the supplemental Methods.

We performed 10 sensitivity analyses. In the first 8 analyses, the study population was rendered more homogeneous by excluding (1) patients who participated in an rFVIII product clinical trial; (2) or by selecting patients with an intron-22 inversion; (3) patients with no family history of inhibitors; (4) patients with white parents; (5) patients who fulfilled criteria 2, 3, and 4; (6) patients born between 2000 and 2009 (as in the RODIN population); (7) only EDs in the 2004-2014 period during which products D and E were both marketed; and (8) patients recruited by the most contributory HTCs to adjust for HTC in the multivariate analysis. In the ninth analysis, only the rFVIII product received at the first infusion (fixed factor) was considered. Finally, the main analysis was repeated by means of pooled logistic regression with the cumulative number of EDs as the time variable, which corresponded to the statistical method used in the RODIN study.²⁰ 

We combined HRs for Product D compared with Product E (D/E) from the RODIN study and our study using indirect log HR and variance estimation²¹  for all inhibitors and high-titer inhibitors outcomes. The degree of inconsistency across both studies was estimated by the I² statistic.²² 

The French hemophilia surveillance system (Suivi thérapeutique National des Hémophiles, known as Réseau FranceCoag from 2003 onward) has been fully supported by the public health authorities since 1994. The additional data collection for the first 75 EDs was partially supported by the French National Clinical Research Program in 2009 and by Assistance Publique-Hôpitaux de Marseille. No public health authority representatives, except authors from the French Institute for Public Health Surveillance (V.G. and V.H.), had a role in the study design, data collection, data analysis, data interpretation, or report writing.

In all, 741 boys with severe HA born between 1991 and 2013 were recruited to the FranceCoag Network by 37 HTCs. The PUP cohort criteria were met by 492 of these boys, 52 (17.6%) of those born from 1991 to 1999 (n = 296), and 440 (98.9%) of those born from 2000 to 2013 (n = 445). Among these 492 eligible patients, 189 were excluded from the analyses: 12 had not started treatment at the last clinical visit, 110 received pdFVIII at the first infusion, 17 had insufficient data for the first EDs, and 50 also participated in the RODIN study. No patients were excluded because of their follow-up in a particular HTC. Finally, 303 patients followed in 33 HTCs were included in the analyses (Figure 1). Among the 6 rFVIII products marketed in France during the study period (1993 to 2014), those most frequently received for the first infusion were Products D (36.6%), E (32.0%), A (15.8%), and C (8.9%). The last 2 products (B and F) were received as first product by only 10 and 10 boys, respectively, and are therefore not presented individually in the Results section. Among the boys’ fixed characteristics (Table 2), only the calendar period at the first infusion was significantly associated with the first rFVIII product received (P < .001), owing to the staggered market release dates of the products. Among the time-varying risk factors, a regular prophylaxis regimen was initiated within the first 50 EDs in 47.9% of patients. At least 1 peak treatment episode lasting at least 3, 5, and 10 consecutive days was experienced by 69.3%, 39.9%, and 13.5% of patients, respectively. Similarly, at least 1 surgical procedure and at least 1 severe bleeding episode were experienced by 12.5% and 11.6% of patients, respectively. Only the initiation of regular prophylaxis within the first 50 EDs was significantly associated with the first rFVIII product received (P = .047 for global test).

The observation period totaled 478 person-years and 14 044 EDs. Follow-up data are shown in supplemental Table 2 according to the first product received. In total, 274 boys (90.4%) received a single rFVIII product throughout follow-up. The other 29 boys switched at least once to another rFVIII product (supplemental Table 3). Most switches were from a first-generation rFVIII product to a second- or third-generation rFVIII product of the same brand. Five of these boys developed an inhibitor. The contributions in EDs according to the rFVIII product received during the follow-up and the 7 studied time-varying risk factors are shown in supplemental Table 4.

In all, 1417 inhibitor assay results were recorded. On average, these assays were performed every 6.3 EDs during the first 25 EDs and every 9.9 EDs during the overall follow-up period. The assay frequency was very similar across the rFVIII products (supplemental Table 5). A clinically significant inhibitor was diagnosed in 114 boys (37.6%) after a median of 13 EDs (interquartile range [IQR], 8-19 EDs) and at a median age of 15.2 months (IQR, 11.1-22.8 months) (Table 3). Among the boys with inhibitors, 95 (83.3%) were subsequently treated with a bypassing agent and/or ITI at any time during FranceCoag follow-up. The cumulative incidence at ED75 was 40.2% (95% confidence interval [CI], 34.8%-46.2%). A high-titer inhibitor was diagnosed in 63 boys (20.8%), with a cumulative incidence of 23.9% (95% CI, 19.1%-29.6%) at ED75 (supplemental Figure 3). A clinically significant inhibitor was diagnosed in only 3 boys between ED50 and ED75, and no cases were diagnosed between ED75 and ED100 among the 303 selected boys.

The risk of inhibitor development was different across the 4 studied rFVIII products for the all inhibitors outcome (P = .025) and also for the treated inhibitors outcome (P = .019) (Table 4 and supplemental Figure 4). Specifically, the inhibitor risk was significantly higher with Product D than with Product E (P = .031 and P = .046 for both outcomes, respectively). Similar tendencies were observed in the multivariate analyses: D/E aHR was 1.55 (95% CI, 0.97-2.49) for all inhibitors, D/E aHR was 1.56 (95% CI, 0.82-2.98) for high-titer inhibitors, and D/E aHR was 1.58 (95% CI, 0.94-2.64) for treated inhibitors (P < .2). No significant difference or stable trend was observed for Product A or C vs Product E.

Detailed results of the 10 sensitivity analyses are shown in supplemental Tables 6-15. Estimated aHRs were between 1.52 and 2.59 for all inhibitors outcomes across the primary analysis and the 10 sensitivity analyses, between 1.29 and 2.00 for the high-titer inhibitors outcomes, and between 1.47 and 2.39 for the treated inhibitors outcomes (Figure 2). Wide HR CIs were observed for Products A and C, particularly in sensitivity analyses with reduced populations, and no stable trend was detected. Finally, the higher risk associated with Product D vs Product E was stable at approximately 60% for the 3 outcomes. P values were lower than .2 for all inhibitors and treated inhibitors outcomes.

The RODIN-FranceCoag combined D/E aHR was 1.58 (95% CI, 1.17-2.14) for all inhibitors and 1.70 (95% CI, 1.15-2.52) for high-titer inhibitors (Figure 3). No heterogeneity was observed between the RODIN and the FranceCoag studies for both outcomes (I² = 0%).

A publicly funded pharmacosurveillance system for antihemophilia drugs has existed in France since 1994. We took advantage of this data set to challenge the unexpected and disturbing finding of the RODIN study.¹⁴  Both RODIN and FranceCoag cohort studies precisely recorded the first EDs of replacement therapy under real-life conditions. The cumulative incidence rates of inhibitor development by ED75 were 32.4% and 40.2% for all inhibitors, and 22.4% and 23.9% for high-titer inhibitors in the RODIN study and our study, respectively.

We observed a significant association between the rFVIII product received and the all inhibitors outcome. More specifically, after taking known major genetic and nongenetic cofactors into account, the D/E aHR was 1.55 (95% CI, 0.97-2.49). The D/E aHR was 1.56 (95% CI, 0.82-2.98) for high-titer inhibitors and 1.58 (95% CI, 0.94-2.64) for treated inhibitors. The events number in our study is lower than that in the RODIN study and leads to a lower analysis power. However, the difference between inhibitor incidence with Products D and E persisted in 10 sensitivity analyses based on more homogeneous subgroups and different statistical approaches. Meta-analysis of the RODIN and FranceCoag results showed a concordant and significant 60% higher risk with Product D compared with Product E for the all inhibitors and high-titer inhibitors outcomes. In contrast, in the literature only 9 (15%) of 60 PUPs and minimally treated patients enrolled in registration trials of Product D developed inhibitors,^23,24  and postmarketing studies showed even lower rates (supplemental Table 21).^25,26  One key difference with cohort studies is that patients are not selected, whereas clinical trials exclude some patients at an increased risk of inhibitor development, such as those requiring intensive initial treatment because of early severe bleeding.

The prevalence of HA at birth in France from 1991 to 2008 was estimated at 23.3 cases per 100 000 male live births (supplemental Table 1). This result, which represents one of the highest rates observed in an industrialized country,²⁷  supports the exhaustiveness of the FranceCoag Network. All but 5 of the boys with severe HA born since 2000 met the enrollment criteria for the PUP cohort. Among the 370 PUPs with severe HA treated with rFVIII, only 17 (4.6%) were excluded because of insufficient data. Our findings are thus representative of PUPs with severe HA treated in France. More than half the boys (51.5%) were enrolled after a few EDs, regardless of whether or not an inhibitor had already been discovered. Their clinical history, including treatments, was fully recorded from birth. Thus, these boys do not correspond to minimally treated patients, and their inclusion does not constitute a selection bias. Another potential selection bias is that patients with known and identifiable risk factors for inhibitor development at treatment outset might have received Product D preferentially. This prescription bias is conceivable from 2002 and 2005, years in which 2 articles reported a notably low rate of inhibitor development (15%) in PUPs treated with Product D.^23,24  During this period only the main genetic risk factors for inhibitor development were identified in PUPs. In our study, slightly higher proportions of patients with a high-risk F8 gene defect and a family history of inhibitors were observed among PUPs treated with Product D than among those treated with Product E, whereas the opposite was observed in the RODIN study.¹⁴  Moreover, owing to the high proportion of sporadic cases among patients with severe HA, the F8 gene defect is often unknown when choosing the FVIII product for initial treatment.²⁸  It is therefore unlikely that such a prescription bias occurred in either cohort. Furthermore, the 3 main genetic risk factors were included in both multivariate analyses. High-titer inhibitors are usually diagnosed after hemorrhagic events or ineffective replacement treatment, and any underreporting would preferentially concern low-titer inhibitors. Such a bias might be envisaged if the frequency and/or sensitivity of inhibitor screening assays varied according to the rFVIII product received. Because the calendar period clearly influenced the choice of product and might have affected the outcome assessment,^7,15  we considered it in multivariate analyses. Furthermore, Products D and E were used during concurrent time periods (supplemental Figure 1). Differences in practices across HTCs with respect to the choice of rFVIII products and outcome assessments are also conceivable, but such variations would have been limited because our study took place in a single country. Furthermore, no difference in inhibitor assay frequency was observed between Product D and Product E (supplemental Table 5). We obtained similar results when we restricted the analysis to patients treated in the 13 most contributory HTCs and when we included the HTC effect in multivariate analysis (supplemental Table 13). Ultimately, similar results were observed after successively removing the data for each of the 13 most contributory HTCs and in the 20 less contributory HTCs (supplemental Table 20). We addressed potential genetic confounding factors by taking most of them into account in multivariate analyses. Although greater precision would be desirable,^16,29  we could compare only 2 classes of F8 gene defects because of limited subgroup sizes. Again, similar results were obtained in patients belonging to the largest homogeneous class (intron-22 inversion) (supplemental Table 7). Other highlighted genetic inhibitor risk factors were HLA genotype^30-32  and polymorphisms in immune regulatory genes.^33,34  In clinical practice, these genetic markers are rarely investigated, and there is no argument to support their possible influence in the choice of FVIII product. Genetic risk factors are thus unlikely to have played a confounding role. The best-known potential nongenetic confounding factors—replacement treatment intensity³⁵  and related conditions (surgical procedures and severe bleeds)—were factored into our multivariate analyses in addition to regular prophylaxis.

The main difference between Product D and Product E is that Product D is produced in baby hamster kidney (BHK) cells and Product E is produced in Chinese hamster ovary (CHO) cells. Changes in BHK manufacturing practices between Product B and Product D might have led to increased immunogenicity,³⁶  but the number of patients treated with Product B in both the RODIN study and in our cohort was too small for comparison with Product D. rFVIII molecules produced by BHK and CHO cells differ in several respects, such as the degree of tyrosine residue sulphation and type of glycosylation.³⁷  The presence of specific glycan chains might affect dendritic cell uptake and thereby modify rFVIII immunogenicity.³⁸  The amino acid sequence of the BHK rFVIII products (B and D) corresponds to the most frequent haplotype (H1) in the white population.³⁹  Full-length CHO rFVIII products (A and E) have an amino acid difference at B-domain position 1241, but no substantial interaction was found between the patient’s F8 haplotype, rFVIII product received, and inhibitor risk.⁴⁰  Furthermore, these products also differ in their biological activities.⁴¹  In fact, the mechanism of the difference in inhibitor incidence between rFVIII remains unclear and has not been elucidated by this study. Nonclinical studies will thus be necessary to identify the mechanism underlying the difference in immunogenicity between Products D and E. However, it is conceivable that a slight difference in immunogenicity among rFVIII products could lead to a moderate increase in the risk of inhibitor development.⁴²  Unfortunately, few if any other HA PUP cohorts exist worldwide, ruling out further epidemiologic results. The consistency between our findings and those of the RODIN study suggests (but does not prove) that the observed association between rFVIII products and the risk of inhibitor development is causal. The potential impact of the higher risk of inhibitor development associated with Product D could be estimated to 1 or 2 extra cases annually in France (supplemental Results). Thus, we think that these results concerning a major adverse effect are sufficiently convincing to warrant consideration in the choice of rFVIII products for PUPs with severe HA in France and in other countries.

American and European medicine agencies recognize that registration trials with small numbers of selected patients may not be able to detect a moderate difference in FVIII immunogenicity. Several initiatives were implemented in the past decade to improve our knowledge of the association between FVIII products and inhibitor incidence. The ongoing Survey of Inhibitors in Plasma-Product Exposed Toddlers (SIPPET) trial was launched in 2009 to compare incidence of inhibitors with von Willebrand factor–containing pdFVIII products versus rFVIII products in PUPs with severe HA.⁴³  The expected number of patients in the rFVIII arm may be too small to demonstrate a difference within this group. To the best of our knowledge, no other randomized trials comparing FVIII products are planned, owing particularly to the rarity of HA and the very young age of the target population. In the United States, the Centers for Disease Control and Prevention began the Hemophilia Inhibitor Research Study in 2006 to determine the feasibility of using a public health surveillance system to collect key information about inhibitors.⁴⁴  This feasibility study has not yet produced results for PUPs with HA, but national implementation of this inhibitor surveillance system is being considered. The European Hemophilia Safety Surveillance System (EUHASS) was established concurrently.⁴⁵  Adverse events and data for exposed populations are periodically reported, but a stable number of patients treated with a particular product over at least a 2-year period is required to estimate the cumulative incidence of inhibitors (at ED50 in PUPs) with that product.⁴⁶  This could limit the precision of product comparisons in this pharmacovigilance program. Our findings and those of the RODIN study highlight the need to strengthen the power and responsiveness of postmarketing monitoring of hemophilia treatments worldwide. This is a major issue because several new FVIII products, including human cell–derived and long-acting rFVIII are reaching an advanced stage of clinical development and may be marketed soon in several countries. Given the heavy burden of inhibitors for individual patients and the high costs of bypassing agents and ITI, this mobilization would not only benefit patients but would also lead to substantial savings for national health care budgets.

The authors thank Dominique Costagliola for critical review of the manuscript and helpful discussions. We also thank Alexandra Doncarli (coordinating center) and clinical research assistants Vincent Dalibard, Yves Guillaume, and Noema Nemausat for their assistance in data collection, as well as all the contributors from the FranceCoag Network for their participation. T.C. thanks Jennifer Markovic for English editing assistance.

Contribution: T.C., H.C., and J.G. designed the FranceCoag Network and wrote the manuscript; T.C. and V.H. performed the statistical analyses; R.d.O., S.C.-D., J.G., and V.R.-R. validated all inhibitor cases; all authors designed the analytical strategy, discussed the results, critically reviewed the manuscript, and had final responsibility for the decision to submit.

Conflict-of-interest disclosure: T.C. has received support for attending scientific meetings from Baxter Bioscience and LFB. H.C. has received support for attending scientific meetings and honoraria (speaker fees/consultant in advisory boards) from Baxter Bioscience, Bayer HealthCare, CSL Behring, LFB, Novo Nordisk, and Pfizer; has been an investigator in studies sponsored by Baxter, Bayer, CSL Behring, LFB, Octapharma, and Pfizer; and has received research support from CSL Behring, LFB, and Novo Nordisk (none of these relate to this study). S.C.-D. has received support for attending scientific meetings and honoraria (speaker fees/consultant in advisory boards) from Baxter Bioscience, Bayer HealthCare, CSL Behring, LFB, Novo Nordisk, and Pfizer and has been an investigator in studies sponsored by Baxter, LFB, and Pfizer. R.d.O. received support for attending scientific meetings and honoraria as consultant or invited speaker from Baxter Bioscience, Bayer HealthCare, CSL Behring, Novo Nordisk, Pfizer, and Swedish Orphan Biovitrum. B.G. has received support for attending scientific meetings and honoraria (speaker fees/consultant in advisory boards) from Baxter Bioscience, Bayer HealthCare, CSL Behring, and Novo Nordisk; has been an investigator in studies sponsored by Baxter Bioscience, Bayer HealthCare, LFB, Novo Nordisk, Octapharma, and Pfizer; and has received research support from Baxter Bioscience, Novo Nordisk, and Pfizer (none of these relate to this study). V.M. has received support for attending scientific meetings from Bayer HealthCare and LFB. C.R. has received support for attending scientific meetings and honoraria (speaker fees/consultant in advisory boards) from Baxter Bioscience, CSL Behring, LFB, Novo Nordisk, Pfizer, and Swedish Orphan Biovitrum; has been an investigator in studies sponsored by Baxter, CSL Behring, LFB, and Pfizer; and has received research support from CSL Behring and Novo Nordisk (none of these relate to this study). V.R.-R. has received support for attending scientific meetings from Baxter Bioscience, Bayer HealthCare, CSL Behring, Novo Nordisk, Octapharma, and Pfizer; and has been an investigator or a co-investigator in studies sponsored by Bayer, LFB, Novo Nordisk, and Pfizer (none of these relate to this study). C.V. has received support for attending scientific meetings and honoraria (speaker fees) from LFB and Pfizer. J.G. has received support for attending scientific meetings and honoraria (speaker fees/consultant in advisory boards) from Baxter Bioscience, Bayer HealthCare, LFB, Novo Nordisk, and Pfizer; has been an investigator in studies sponsored by Baxter, LFB, Novo Nordisk, and Pfizer; and has received research support from Novo Nordisk (none of these relate to this study). The remaining authors (V.G. and V.H.) declare no competing financial interests.

A complete list of contributors to the FranceCoag Network appears in supplemental Material.

Correspondence: Thierry Calvez, Institut Pierre Louis, INSERM et Université Pierre et Marie Curie, Unité Mixte de Recherche en Santé 1136, 56 Boulevard Vincent Auriol, CS 81393, 75646 Paris Cedex 13, France; e-mail: thierry.calvez@ccde.chups.jussieu.fr.