In non-severe hemophilia A the risk of inhibitor after intensive factor treatment is greater in older patients: a case–control study


Christine L. Kempton, 2015 Uppergate Drive, Atlanta, GA 30322, USA.
Tel.: +1 404 7272846; fax: +1 404 7273681.


Summary. Background: Twenty-five percent of new anti-factor VIII (FVIII) antibodies (inhibitors) that complicate hemophilia A occur in those with mild and moderate disease. Although intensive FVIII treatment has long been considered a risk factor for inhibitor development in those with non-severe disease, its strength of association and the influence of other factors have remained undefined. Objective: To evaluate risk factors for inhibitor development in patients with non-severe hemophilia A. Methods: Information on clinical and demographic variables and FVIII genotype was collected on 36 subjects with mild or moderate hemophilia A and an inhibitor and 62 controls also with mild or moderate hemophilia A but without an inhibitor. Results: Treatment with FVIII for six or more consecutive days during the prior year was more strongly associated with inhibitor development in those ≥ 30 years of age compared with those < 30 years of age [adjusted odds ratio (OR) 12.62; 95% confidence interval (CI), 2.76–57.81 vs. OR 2.54; 95% CI, 0.61–10.68]. Having previously received < 50 days of FVIII was also not statistically associated with inhibitor development on univariate or multivariate analysis. Conclusions: These findings suggest that inhibitor development in mild and moderate hemophilia A varies with age, but does not vary significantly with lifetime FVIII exposure days: two features distinct from severe hemophilia A.


Although persons with severe hemophilia A [factor VIII (FVIII) activity < 1%] are at greatest risk of developing an anti-FVIII antibody (inhibitor), up to 25% of new inhibitors develop in patients with non-severe disease (FVIII activity 1–40%) [1]. Most of our knowledge of risk factors for inhibitor development in those with non-severe hemophilia A has come from uncontrolled observations or single institutional studies. These studies have reported an association between intensive FVIII treatment and inhibitor formation. The influence of factors such as age, FVIII genotype, number of lifetime FVIII exposure days and race on inhibitor formation are less well understood in this population. The largest reported case series in non-severe hemophilia A included 26 subjects of whom 16 (61.5%) developed an inhibitor after intensive FVIII replacement therapy for surgery, trauma or muscle bleeding [2]. In a recently reported longitudinal observational cohort study from a single institution, receiving intensive FVIII treatment for the first time while undergoing surgery carried a relative risk (RR) of inhibitor formation of 186 [95% confidence interval (CI), 25–1403] [3].

The most frequently reported FVIII gene (F8) missense mutation associated with inhibitors in non-severe hemophilia A is R593C [2,4]. Other commonly reported F8 missense mutations include W2105C [5], R2150H [6] and W2229C [2]. Overall, F8 missense mutations reported in association with inhibitors in non-severe hemophilia A have clustered in one of three regions, leading some to postulate that mutations within certain sequences in the A2 and A3 domains and near the junction of the C1 and C2 domains are more immunogenic [2,7,8]. Only one study has evaluated the association of the FVIII genotype with inhibitor formation in persons with non-severe hemophilia A [3]. Although the previous study found the R593C mutation to be associated with inhibitor formation [relative risk (RR) 10, 95% confidence interval (CI) 0.9–119], the population studied was enriched with this mutation (38% of subjects).

The present study was undertaken in a diverse United States population to further evaluate the association of intensive FVIII treatment and inhibitor formation in non-severe hemophilia A and to identify additional risk factors for inhibitor formation, including the FVIII genotype.


Study subjects

This study, a case–control design, was approved by Institutional Review Boards at all participating sites. Subjects were enrolled during an 18-month period beginning July 2007 and ending December 2008.

Case selection.  Cases were defined as individuals with mild or moderate hemophilia A (FVIII 1–40%) based on local FVIII testing and a history of either two inhibitor titers ≥ 1 BU mL−1 or one such inhibitor titer followed by the initiation of immune tolerance. As a first step in case identification, 4653 males with non-severe hemophilia A that comprise the Universal Data Collection (UDC) data set compiled by the Division of Blood Disorders of the Centers for Disease Control and Prevention (CDC) [9] were screened for a history of one elevated inhibitor titer. In order to initially be inclusive, only one positive inhibitor titer was used as screening criteria. From this screen, 110 males with mild or moderate hemophilia A were identified at 58 hemophilia treatment centers (HTCs) (Fig. 1). All HTCs with a potential case subject were contacted to determine if the subject met inclusion criteria as a true case and, if so, the HTC was invited to participate in the study. Of the 110 potential cases initially identified on screening, 30 individuals at 24 HTCs were verified to meet the case definition. Of the 24 HTCs with cases, 16 participated and enrolled 13 of the cases identified in the initial screen. After reviewing their own patient records, which included patients who developed an inhibitor prior to the start of UDC data collection in 1998 and patients not enrolled in the UDC project, the 16 participating HTCs identified and enrolled an additional 30 cases. The 13 case subjects originally identified from the UDC cohort and the additional 30 subjects identified by participating centers comprised the 43 enrolled case subjects. Of the potential case subjects not included in the study, three were verified by participating HTCs, but were not enrolled; 14 were verified to meet the inclusion criteria, but the HTC declined to participate; 66 did not meet the study inclusion criteria because they had severe disease, did not have an inhibitor or the inhibitor titer was ≥ 1 BU mL−1 on only one occasion without initiation of immune tolerance; and 14 could not be verified because of a lack of response to queries. Information on race and baseline FVIII levels was available on the 17 subjects that were verified to meet inclusion criteria, but were not enrolled.

Figure 1.

 Flow diagram of case subject enrollment. UDC, universal data collection; BU, Bethesda Unit; HTC, hemophilia treatment center; ITI, immune tolerance induction.

Control selection.  Each participating center was asked to enroll two control subjects for each case that the center enrolled in the study. Only males with mild or moderate hemophilia A (FVIII 1–40%) with prior exposure to FVIII but no history of an inhibitor were eligible to participate as controls. Any prior exposure to FVIII was used as an inclusion criterion for control subjects to include only those that would be considered at risk for inhibitor development. To ensure that only truly inhibitor negative controls were included, only patients with previous inhibitor titers < 0.6 BU mL−1 were eligible. All control subjects were confirmed to have an inhibitor titer of < 0.6 BU mL−1 by testing at the CDC Molecular and Hemostasis Laboratory using blood samples collected at enrollment. Using these criteria, 66 control subjects were enrolled (1.5 control subjects per case subject).

Data collection

After written informed consent was obtained, staff at HTCs completed a standard data collection form for each case and control subject. The primary exposure of interest was intensive FVIII treatment, defined as six or more consecutive days of FVIII replacement. Data were collected on intensive FVIII treatment during the year preceding inhibitor development for cases and during the year preceding enrollment for controls. Where possible, information was based on medical record review, but if not available, patient recall was employed. Other data collected included the following: self-identified race (Black or African American, White, Asian or other); ethnicity (Hispanic or non-Hispanic); family history of hemophilia and inhibitor; age at first factor infusion; HIV and hepatitis C antibody (positive, negative and unknown); the type of FVIII products used; lifetime total number of FVIII exposure days prior to inhibitor detection in cases or prior to enrollment in controls; and FVIII genotype if known. In those that received intensive FVIII treatment, additional data were collected as follows: method of FVIII delivery (bolus injection only or any period of continuous infusion); highest daily dose; duration of daily FVIII treatment; indication for treatment (joint bleed – spontaneous or traumatic, muscle bleed – spontaneous or traumatic, bone fracture, surgery, intracranial hemorrhage or other); type of FVIII product used during the intensive treatment; duration of intensive FVIII treatment and if infection (positive culture or treatment with antibiotics) complicated the period of intensive FVIII treatment.

Laboratory materials and methods

Blood was collected via venipuncture into two 4.5-mL vacuum-sealed tubes containing 3.2% sodium citrate. Platelet-poor plasma was obtained by centrifugation. Red cells and buffy coat were collected for DNA extraction. Samples were either stored at 2–8 °C and shipped to the Molecular and Hemostasis Laboratory at the CDC within 24 h or stored at −70 °C until shipment after 24 h.

Inhibitor titers were measured on control subject samples at CDC by a modification of the Nijmegen–Bethesda assay [10]. An inhibitor titer ≥ 0.6 BU mL−1 was considered positive.

FVIII genotyping was performed by the Molecular and Hemostasis Laboratory at the CDC except when historical FVIII genotype data were available (historical FVIII genotype was used in three subjects). All exons, intron–exon junction regions and the 3′ untranslated regions of F8 were sequenced in both directions by automated sequencer. The VariantSEQr™ protocol was used for resequencing on a 3730 DNA Analyzer from Applied Biosystems (Carlsbad, CA, USA). The PCR primers and M13 sequencing primers are described at with a few modifications to the PCR primers to enhance throughput and reproducibility. Data were analyzed with SeqScape® (Applied Biosystems). Inversions of intron 22 and intron 1 in F8 were examined by PCR [11]. Mutations were assigned Human Genome Variation Society (HGVS) names for protein amino acid changes using a Mutalyzer sequence variation nomenclature checker at from reference sequences NM_000132.3 and NP_000123.1.

Missense mutations were classified according to whether the mutation was within specific regions of the A2 or A3 domains or C1/C2 domain (amino acids 535–663, 1854–2019 and 2020–2286, respectively)[8].

Data analysis

Distribution of characteristics was compared for cases and controls using the χ2 distribution. For sparse data, Fisher’s exact test was utilized to calculate P-values. Age and baseline FVIII activity was compared between cases and controls using Student’s t-test. The odds ratio (OR) was calculated from a 2 × 2 table and 95% confidence intervals (CI) were calculated. Statistical significance of the OR was assessed using the χ2-test. To adjust for confounding, a stratified analysis was performed. When the OR was relatively constant between subgroups, it was combined using the Mantel–Haenszel method to form an adjusted OR. Confounding was considered present if the adjusted OR varied by 10% or greater from the unadjusted OR. The presence of effect modification was assessed using the Breslow Day test for heterogeneity. In addition to a stratified analysis, multivariate analysis was done using unconditional logistic regression. Predictor variables were included if they were potential confounders, demonstrated effect modification in the stratified analysis, have been associated with inhibitors in severe hemophilia A or were potentially independently associated with the outcome, inhibitor formation. For univariate and multivariate analyzes baseline FVIII level and patient age were made into categorical variables. Because of the increased risk of inhibitor formation with low FVIII levels, baseline FVIII activity was categorized into two groups, 1 to < 2% or ≥ 2%. Age was arbitrarily categorized into two equal groups defined by the median age of study subjects, < 30 or ≥ 30 years. The statistical significance of interaction terms in the multivariate model, further indication of effect modification, was assessed using both the Wald test and likelihood ratio test (LRT).

The subgroup of cases and controls exposed to intensive FVIII treatment were analyzed. Cases and controls were compared using Wilcoxon’s rank sum test for continuous variables or Fisher’s exact test for categorical variables.

Twenty subjects (eight cases and 12 controls) comprised 10 pairs of related subjects (one pair of twins, six pairs of brothers and three pairs of cousins). Related pairs were concordant for the presence of an inhibitor in the pair of twins and one pair of brothers and concordant for the absence of an inhibitor in four pairs of brothers. The remaining four pairs were discordant. The possible influence of these related pairs on variance estimates was investigated by assessing the variance of the parameter estimate for the association between the outcome and predictor variables after performing 1000 simulations in which one member of each of the 10 related pairs was selected at random for inclusion in the analysis. Based on these analyzes, the influence of the related pairs on the variance was small therefore all related pairs were kept in the study population. All analyzes were performed using sas version 9.2 (Cary, NC, USA).


Although genotype was not included in the initial inclusion/exclusion criteria, several subjects were demonstrated to have an intron-22 inversion. In order to prevent misclassification, subjects with mutations previously reported to be associated only with severe disease in the HAMSTeRS database (accessed June 2009) were excluded from the analysis (four cases and two controls with intron-22 inversion, one case with nonsense mutation R427Stop, one case with nonsense mutation W2046Stop, one case with deletion exon 1 and one control with delA1194) [12]. In addition, one control subject was excluded because a negative inhibitor titer at enrollment could not be verified by inhibitor testing at the CDC. The remaining 36 cases and 62 controls had complete data and formed the study population. All 62 control subjects included in the analysis had a negative inhibitor titer at the time of enrollment [median inhibitor titer 0.1 BU mL−1 (range 0–0.3 BU mL−1)]. When the potential cases that were verified, but not enrolled, were compared with enrolled cases, there was no difference in FVIII levels. Among the enrolled case subjects, the inhibitor titer became positive after the year 2000 in 24 case subjects (66%), during the 1990s in 10 subjects (27%), during the 1980s in one subject (2.8%) and during the 1970s in one subject (2.8%). Four case subjects (11.1%) had started immune tolerance after only one positive inhibitor titer. Seventy-five percent of case subjects had a maximum inhibitor titer > 5 BU mL−1, including thre of the four that started immune tolerance after one positive titer.

Neither the mean age nor the baseline FVIII level was different between cases and controls [31.0 years vs. 31.1 years (P = 0.94) and 7.6% vs. 8.1% (P = 0.73)]. The baseline and exposure characteristics of cases and controls are shown in Tables 1 and 2. Cases were more likely to have received intensive FVIII treatment compared with controls (P < 0.001). Distributions of age, race, ethnicity, family history of an inhibitor, total number of FVIII exposure days and type of product used were not different between case and control subjects. Of the 17 case subjects that were verified but not enrolled, 22% were of black race. If these potential case subjects had been enrolled, the proportion of black case subjects would have increased to 13.2% but the distribution of race amongst cases and controls would remain similar (P = 0.54).

Table 1.   Frequency of characteristics in groups
CharacteristicsCases n = 36Controls n = 62Chi-squareP-value
N (%)N (%)
  1. *Fisher’s exact test. FVIII, factor VIII; HIV, human immunodeficiency virus; HCV, Hepatitis C virus.

Intensive FVIII treatment18 (50.0)11 (17.7)11.38< 0.001
 < 30 years17 (47.2)33 (52.4)0.240.62
 30–60 years13 (36.1)24 (38.7)  
 > 60 years6 (16.7)5 (8.1)  
 White32 (88.9)56 (90.3)0.160.92
 Black3 (8.3)5 (8.1)  
 Other1 (2.6)1 (1.6)  
Hispanic ethnicity2 (5.6)8 (12.9)0.32*
Family history of inhibitor7 (19.4)4 (6.5)0.09*
Baseline FVIII
 1–< 2%5 (13.9)4 (6.4)2.080.35
 2–5%16 (44.4)25 (38.7)  
 > 5%15 (41.7)33 (53.2)  
Age at first factor infusion
 ≤ 2 years10 (27.8)23 (37.1)3.280.19
 3–10 years11 (30.6)24 (38.7)  
 > 10 years15 (42.7)15 (24.2)  
Prior FVIII exposure
≤ 20 days6 (16.7)15 (24.2)5.750.12
21–50 days15 (41.7)12 (19.4)  
51–100 days5 (13.9)13 (20.9)  
> 100 days10 (27.8)22 (34.5)  
Product during prior year
 Plasma-derived8 (22.9)9 (14.5)3.860.14
 Recombinant25 (71.4)41 (66.1)  
 None2 (5.7)12 (19.4)  
 Negative32 (88.9)58 (93.6)0.660.72
 Positive2 (5.6)2 (3.2)  
 Unknown2 (5.6)2 (3.2)  
 Negative21 (58.3)38 (61.2)2.510.29
 Positive11 (30.6)22 (35.4)  
 Unknown4 (11.1)2 (3.2)  
Table 2.   Frequency of factor VIII mutation in groups
N (%)N (%)
  1. *Fisher’s exact test. fVIII, factor VIII.

Type of fVIII mutationN = 36N = 62  
 Missense32 (88.9)53 (85.5)5.120.16
 Deletion2 (5.56)0  
 Splice site1 (2.8)5 (8.1)  
 None detected1 (2.78)4 (6.45)  
Missense mutations in regions previously linked to inhibitorN = 32N = 53  
 A2 (535–663)8 (25.0)10 (18.9)1.230.75
 A3 (1854–2019)8 (25.0)10 (18.9)  
 C1/C2 (2020–2286)5 (15.6)10 (18.9)  
 None11 (34.4)23 (43.4)  
Domain location of missense mutation
 A1 (1–372)3 (9.4)3 (5.66)2.690.61
 A2 (373–740)11 (34.4)22 (42.5)  
 A3 (1690–2019)11 (34.4)14 (25.4)  
 C1 (2020–2172)2 (6.3)8 (15.1)  
 C2 (2173–2332)5 (15.6)6 (11.3)  
Specific missense mutations
 R593C7 (19.4)3 (4.9)0.03*
 N1922S6 (16.7)4 (6.5)0.16*
 R2150H3 (8.3)3 (4.9)0.67*
 W2229C01 (1.6)

There was no association between inhibitor development and the presence of a missense mutation grouped within specific regions of the A2, A3 or C1/C2 domain of F8 where missense mutations have been previously reported to commonly occur (Table 2). In both cases and controls the majority of missense mutations were in the A2 and A3 domains. The missense mutation R593C was significantly more common in cases than controls (19.4% vs. 4.8%, P = 0.03). The missense mutation N1922S was found more commonly in cases compared with controls (16.7% vs. 6.5%), but the difference was not statistically significant (P = 0.16).

The univariate association between these characteristics and inhibitor formation is shown in Table 3. Having received intensive FVIII treatment during the prior year was strongly associated with inhibitor development (OR 4.64).

Table 3.   Univariate association of characteristics with inhibitor development
CharacteristicsOdds ratio95% CI
  1. *Use of a recombinant product vs. the combination of plasma-derived and no product use. FVIII, factor VIII.

Age < 30 years0.780.35–1.79
Race: White0.860.23–3.27
Baseline FVIII 1 to < 2%2.340.59–9.34
Family history of inhibitor3.500.95–12.93
< 50 Lifetime exposure days to FVIII1.820.79–4.17
Intensive FVIII treatment4.641.84–11.67
Recombinant product during the prior year*1.160.48–2.81
Age ≤ 5 years at first factor exposure0.840.34–1.91
R593C missense mutation4.751.14–19.71
N1922S missense mutation2.900.78–11.07

The association between intensive FVIII treatment and inhibitor formation was further explored utilizing a stratified analysis. Age (< 30 years or ≥ 30 years) showed effect modification on the association between intensive FVIII treatment and inhibitor formation (Breslow-Day P = 0.03). The association between intensive FVIII treatment and inhibitor formation was greater in those 30 years of age or older (OR 13.54) compared with those < 30 years of age (OR 1.55). After adjustment for other variables (white race, baseline FVIII activity 1 to < 2%, < 50 lifetime days of FVIII exposure prior to inhibitor formation or enrollment, < 5 years of age at first FVIII exposure and recombinant product use), no confounding or effect modification was seen.

On multivariate analysis (Table 4), intensive FVIII treatment was strongly associated with inhibitor development in those 30 years of age or older and not in those < 30 years of age (test for interaction LRT P = 0.12 and Wald P = 0.13). FVIII activity of 1 to < 2% demonstrated a trend toward a statistically significant association with inhibitor formation (P = 0.06). The R593C missense mutation was no longer statistically significantly associated with inhibitor formation. Having < 50 prior FVIII exposure days was also not statistically significantly associated with inhibitor formation. None of the following variables were associated with inhibitor formation and, therefore, were excluded in the final model: a family history of inhibitor, having received a recombinant FVIII product, having received FVIII for the first time prior to 5 years of age or the missense mutation N1922S. When the analysis was limited to those with mild disease only (FVIII > 5%), the results were similar, although the interaction between age and intensive exposure was stronger (age ≥ 30 years OR 39.18 and age < 30 years OR 1.14; LRT for interaction P = 0.05).

Table 4.   Multivariate analysis of subject characteristics and their association with inhibitor formation
CharacteristicsParameter EstimateSEORCI
  1. Hosmer and Lemeshow goodness-of-fit = 2.56, degrees of freedom (DF) = 6, P-value = 0.86. Likelihood ratio = 23.38, DF=7, P-value = 0.002. −2 Log Likelihood = 128.88 for intercept only and 105.49 for intercept and covariates. FVIII, factor VIII. All variables are adjusted for all other variables.

Intensive FVIII treatment
 < 30 years of age0.960.742.540.61–10.68
 ≥ 30 years of age2.540.7812.632.76–57.81
< 50 previous days of FVIII0.450.612.080.74–5.90
Baseline FVIII 1 to < 2%1.590.854.910.92–25.95
White race0.190.781.200.26–5.52

On analysis of the subgroup of subjects (18 cases and 11 controls) who were exposed to intensive FVIII treatment, the median age in cases was 42.5 years [interquartile range (IQR) 38 years] compared with 22 years (IQR 23 years) in controls (P = 0.23). The median baseline FVIII activity was similar between cases and controls [4.5% (IQR 9%) vs. 5.0% (IQR 6%) (P = 0.86)]. The median duration of treatment days was significantly longer in cases compared with controls [14.5 days (IQR 9 days) vs. 10 days (IQR 6 days) (P = 0.03)]. Continuous infusion of FVIII was used during the period of intensive FVIII treatment in 18.2% of controls and 38.9% of cases (P = 0.41). Surgery was the indication for intensive FVIII treatment in 77.8% of cases compared with 36.4% of controls (P = 0.05). Of the 18 subjects (14 cases and four controls) who had surgery as the indication for intensive FVIII treatment, seven received FVIII by continuous infusion and 11 received FVIII by bolus injection only. Having surgery as the indication for intensive treatment was weakly associated with receiving > 50 Units kg−1 day−1 of FVIII but was not associated with longer duration of treatment (> 14 days) (P = 0.06 and 0.47, respectively). In those over 30 years of age, surgery was more likely to be the indication for intensive FVIII treatment compared with those < 30 years (P = 0.01). Of those that underwent surgery as the indication for intensive treatment, 78.6% (11/14) were for orthopedic indications in those over 30 years of age compared with 25.0% (1/4) in those < 30 years of age. Surgical events were complicated by infection in two cases. Joint or muscle bleeding was the indication for intensive FVIII treatment in four case subjects and five control subjects. In the two remaining control subjects, the indication for intensive FVIII treatment was pneumonia requiring a chest tube and pericarditis, respectively. Infection complicated periods of intensive FVIII treatment for non-surgical indications in two controls. In those that developed an inhibitor after intensive FVIII treatment, the inhibitor occurred within 12 weeks of intensive FVIII treatment in 94% (17/18) of cases.


Inhibitor formation in non-severe hemophilia A is a major complication, transforming a manageable disease to one with substantial morbidity. Although much has been learned over the past decade about risk factors for inhibitor formation in those with severe hemophilia A, very little is known about risk factors for inhibitor formation in non-severe hemophilia A. In addition to confirming that intensive exposure to FVIII is a risk factor for inhibitor development, this is the first study to identify an interaction between age and intensive FVIII treatment. This interaction was present even after adjustment for a lifetime exposure to FVIII of < 50 days. Therefore, the impact of intensive treatment in adults does not appear to be the result of either less FVIII exposure prior to adulthood nor because intensive FVIII treatment is their first exposure to FVIII.

Intensive FVIII treatment occurred relatively equally in those < 30 years of age and in those ≥ 30 years of age (24% and 35%, respectively, P = 0.22), however, the indication for intensive FVIII treatment in each age group was different. In those cases that were 30 years of age and older, surgery was the indication for intensive FVIII treatment in 14 out of the 17 subjects with 78.6% of these surgeries for orthopedic indications. Both severe traumatic bleeding and surgery require intensive FVIII treatment and have been hypothesized to potentiate inhibitor development through signals that cause antigens to be perceived as foreign and dangerous thereby promoting an antibody response [13]. If danger signals are considered to be important for inhibitor development in mild and moderate hemophilia A, then these results may indicate that in patients < 30 years of age there are mechanisms for danger signals to occur that are unrelated to intensive FVIII treatment, such as infections or immunizations. Alternatively, danger signals may be less pathophysiologically important for inhibitor development in young persons with non-severe hemophilia A. Although the present study is the first to report on the interaction between age and intensive FVIII treatment, a similar pattern was seen in the cohort reported by Eckardht et al. [3]. In their study, 5 out of the 10 subjects with an inhibitor were older than 30 years of age. Inhibitor development was associated with intensive FVIII treatment in 100% of older subjects and only 40% of younger subjects. In the present study inhibitor development was associated with intensive FVIII treatment in 68% of older subjects and 29% of younger subjects. Although we can speculate that the risk of inhibitor development associated with intensive exposure may be strongly influenced by the indication for intensive treatment and the types of surgeries performed in older patients (i.e. orthopedic surgery), it is clearly multifactorial and this study was not designed to evaluate the independent effect of surgery or a specific type of surgery.

There have been previous reports of the R593C mutation and its association with inhibitor formation in patients with mild hemophilia A [14]. Most recently, in a cohort of subjects from the Netherlands, 8 out of 10 subjects with an inhibitor had the R593C mutation [3]. However, because this mutation was determined to be a founder effect it may have been over-represented in this population. In our more diverse US population, the missense mutation R593C was again associated with inhibitor development on univariate analysis but not multivariate analysis. In the present study, the R593C mutation was seen in subjects from five different states and only two of the ten subjects were known to be related. The N1922S mutation was the second most common mutation seen. Outside of these two specific missense mutations, there did not appear to be a location on the FVIII gene where mutations clustered, as has been previously hypothesized [8].

It is remarkable that 41.7% of case subjects developed their inhibitor after more than 50 prior days of FVIII exposure (Table 1). In severe hemophilia A, inhibitors develop after a median of 9–14 FVIII exposure days [15,16]. In the cohort study reported by Eckhardt et al., [3] the total cohort of 138 mild and moderate hemophiliacs was significantly less treated than our population with a median of 10 FVIII exposure days (IQR 23 days). Accordingly, the effect of more or < 50 prior FVIII exposure days on inhibitor development could not be assessed. The present study cannot exclude any association between prior FVIII exposure and risk of inhibitor development, as it is not adequately powered to detect weak associations. However, it does not appear that the risk of inhibitor development is as significantly different before and after 50 days of FVIII exposure in persons with mild and moderate hemophilia A as it is in those with severe hemophilia A.

The subset analysis suggested that among those who received intensive FVIII treatment, both surgery and longer durations of FVIII treatment were associated with inhibitor formation. Continuous infusion as a method of delivery of intensive FVIII treatment was not significantly associated with inhibitor development in the present study. Although a small effect of continuous infusion cannot be excluded, the strength of the association is less than that between surgery and inhibitor formation. This result is in contrast to the study by Sharathkumar et al. [7]who reported that continuous infusion was associated with inhibitor formation. Additionally, in the cohort study reported by Eckhart et al. continuous infusion was significantly associated with inhibitor formation after adjusting for the R593C missense mutation, receipt of ones first intensive FVIII treatment at the time of surgery and FVIII product change (RR 13; 95% CI 1.9–86). The different results may be due in part to the different study populations: our subjects were older when compared with those studied by Sharathkumar et al. and had more prior FVIII treatment compared with those studied by Eckhardt et al. In addition, we were limited to a very small subset of our population for those analyzes and our findings should be considered preliminary.

There are several limitations of this study. First, the sample size is small leading to a lack of power to detect weak associations. Second, the independent effect of surgery or continuous infusion on inhibitor formation could not be evaluated because these variables were only determined in subjects that received intensive FVIII treatment. Third, the inclusion of related subjects could impair the ability to detect the independent influence of FVIII genotype vs. other genetic modifiers. However, only one related-pair of subjects had the N1922S and the R593C mutation, respectively. Family history of inhibitor was the only variable that was influenced by the choice of which member of a related pair was included in the analysis. As a result, the univariate association between family history of an inhibitor and inhibitor formation could be inflated. As the multivariate model was not altered significantly with the inclusion or exclusion of the variable family history of inhibitor, we felt that inclusion of the related subjects without additional adjustment was appropriate. Fourth, we enrolled subjects based on their reported FVIII activity but their inclusion was re-evaluated when several intron-22 inversions were detected. In order to have a cohort of subjects that clearly represented non-severe hemophilia A, 10 subjects with the FVIII genotype associated only with severe disease in the HAMSTeRS database were excluded from the analysis [12]. However, when these 10 were included, the association with intensive FVIII treatment was not substantially changed (OR, 4.29; 95% CI, 1.81–10.25). Lastly, the number of FVIII exposure days in one-third of subjects was based predominantly on patient interview and recall. However, the use of recall alone was equal amongst cases and controls (33.3% and 35.5%, respectively) and the distribution of prior exposure days was similar amongst those that used patient recall and those that did not regardless of case–control status. Therefore, we do not believe that a significant source of bias was introduced with the use of patient recall to estimate prior FVIII exposure days when complete medical records or calendars were not available.

Overall, the present study has demonstrated that intensive FVIII treatment is strongly associated with inhibitor development in patients with non-severe hemophilia A. The risk is greatest in those 30 years of age or older. Importantly, the risk of inhibitor development does not appear to be as significantly influenced by the number of prior FVIII exposure days as seen in patients with severe disease. Further clinical investigation is needed to understand the effect of age on the risk of inhibitor development associated with intensive FVIII treatment, and what aspects surrounding intensive FVIII treatment promote inhibitor formation. Specifically, it will be important to determine which factors exert the greatest influence: (i) surgical indications and if so, what types of surgery, (ii) method of FVIII delivery, (iii) dose of FVIII, or (iv) duration of FVIII treatment.


C. L. Kempton participated in planning, data collection, data analysis, and writing the manuscript. J. M. Soucie and T. C. Abshire participated in planning, data analysis, and writing the manuscript. C. H. Miller and C. Hooper performed inhibitor assays and genotyping (respectively) and critically reviewed and edited the manuscript. M. A. Escobar, A. J. Cohen, N. S. Key, and A. R. Thompson participated in data collection and critically reviewed and edited the manuscript. All authors had access to the primary data.


The authors wish to acknowledge the following investigators and institutions for their enrollment of subjects in the study: Geoff Allen, Children’s Memorial Hospital, Chicago, IL; Cynthia Gauger, Nemours Children’s Clinic, Jacksonville, FL; John Hord, Children’s Hospital of Akron, Akron, OH; Rebecca Kruse-Jarres, Tulane University, New Orleans, LA; Marilyn J. Manco-Johnson, MD, University of Colorado Denver and The Children’s Hospital, Aurora, CO; Deanna Mitchell, Helen DeVos Children’s Hospital, Grand Rapids, MI; Mark T. Reding, University of Minnesota, Minneapolis, MN; Christopher Walsh, Mount Sinai Medical Center, New York, NY; Guy Young, Childrens Hospital Los Angeles, Los Angeles, CA. We would also like to thank the Hemophilia Treatment Center Network for their stewardship of the UDC project and review of the manuscript, all of the nursing and research staff for the hard work in completing the regulatory documents, enrolling subjects, and completing the data collection, and the CDC laboratory staff for their work on the Bethesda assay and factor VIII genotyping. The study was supported by a grant from CSL Behring Foundation for Advancement of Patient Health and the Cooperative Agreement Prevention of Bleeding Disorder Complications through Regional Hemophilia Treatment Centers. The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.

Disclosure of Conflicts of Interests

C.L. Kempton and A.R. Thompson are consultants for Ipsen Biomeasure. T.C. Abshire is a member of advisory boards for Bayer Healthcare, CSL Behring Foundation, NovoNordisk and Talecris.