Even though technological progress has provided safe therapeutic products for patients with haemophilia A, the development of inhibitors to factor VIII (FVIII) remains a major complication of therapy, making the treatment of bleeding problematic and prophylaxis impossible in many cases. Inhibitors occur in approximately 30% of patients with severe haemophilia A (Addiego et al, 1993; Lusher et al, 1993; Bray et al, 1994), less frequently in those with mild/moderate haemophilia. Inhibitor development is still considered an unpredictable event, although FVIII gene mutations leading to the complete absence of the protein (large deletions, non-sense mutations and intron 22 inversions) are associated with a higher prevalence of inhibitors (Schwaab et al, 1995; Kemball-Cook et al, 1998). Other genetic determinants are suggested by the finding of higher inhibitor incidences in African-Americans and Hispanics (Astermark et al, 2001), and in haemophilic siblings compared with other haemophilic relatives (Cox-Gill, 1999). Mutations and polymorphisms of proteins involved in immune response and cytokine production might influence inhibitor formation, but the results of studies investigating the link between major histocompatibility complex (MHC) class I/II genotype and inhibitor risk were inconclusive (Hay et al, 1997; Oldenburg et al, 1997). On the other hand, that environmental factors influence genetic predisposition to inhibitor formation is suggested by the observation of monozygotic haemophilic twins discordant for inhibitors (European Study Group of Factor VIII Antibody, 1979). Potential environmental factors are the different types of FVIII products used for replacement therapy, but the relative risks of recombinant and plasma-derived FVIII products in inhibitor development are still a matter of debate. Other therapy-related factors potentially involved in the development of inhibitors are mode of treatment delivery by continuous infusion (Sharathkumar et al, 2003), high-dose regimens (Sharathkumar et al, 2003) and very young age at treatment onset (Lorenzo et al, 2001; van der Bom et al, 2003). No data are available on other putative environmental factors for inhibitor formation, such as FVIII treatment in the context of severe bleeding or surgery and during infections, vaccinations or other challenges of the immune system. With this background, this multicentre case–control study was designed to investigate the interactions between environmental factors and FVIII gene defects in a cohort of children with severe or moderately severe haemophilia A who developed inhibitors.
This case–control study investigated the interactions between genetic and environmental factors and inhibitor development in 108 children with haemophilia A exclusively treated with recombinant factor VIII (FVIII). Sixty patients with inhibitors were compared with 48 inhibitor-free controls. Family history of inhibitors and null mutations in the FVIII gene were more prevalent in cases than in controls (20% vs. 2%, P = 0·001 and 83% vs. 64%, P = 0·04, respectively). On the other hand, there was no difference between cases and controls for such putative risk factors of inhibitor development as amniocentesis/villocentesis, premature/caesarean birth, breast-feeding, treatment during infections/vaccinations, surgical procedures and central nervous system bleeding. A trend was found for an increased risk of inhibitor development in children first treated at a young age (<11 months); however, this was not confirmed after adjusting for genetic factors. The implementation of prophylaxis was evaluated as a putative risk factor in a subgroup of 25 cases: seven who started prophylaxis prior to inhibitor development and 18 potentially eligible for prophylaxis because they were inhibitor-free up to the age of 35 months (i.e. the upper limit of the age range at prophylaxis onset in cases and the median age at prophylaxis onset in controls). Patients who started prophylaxis had a lower inhibitor risk than those treated on demand (adjusted odds ratio 0·2, 95% confidence interval 0·06–0·9). The protective effect on inhibitor development shown by prophylaxis may represent an additional advantage prompting its use in haemophilic children.
Eight Italian Haemophilia Centres took part in the study enrolling at least one case and one control, chosen according to the following definitions.
Cases were patients with severe (FVIII <1%) or moderately severe (FVIII: 1–2%) haemophilia A younger than 18 years who developed inhibitors after exclusive treatment with recombinant FVIII (rFVIII) products. Measurable inhibitor titres [>0·5 Bethesda units (BU)/ml] had to be confirmed on at least two occasions. Low-response inhibitors were defined as those with an inhibitor titre persistently <5 BU/ml despite repeated challenges with FVIII concentrates. High-response inhibitors were those with an inhibitor titre >5 BU/ml at any time. Inhibitors were considered transient if they disappeared spontaneously after at least two consecutive positive detections. Controls were patients with severe or moderately severe haemophilia A younger than 18 years treated exclusively with rFVIII products who remained inhibitor-free after more than 50 exposure days (EDs) to factor replacement.
All the participating Centres used the Bethesda method and had introduced the Nijmegen modification (Verbruggen et al, 1995) in the late 1990s. Eligible patients had to be tested for inhibitors at least every 3 months during the first 100 EDs, every 6 months up to 200 EDs and yearly afterwards.
Informed consent to the study was obtained from the parents of 121 patients who met the definition of cases (n = 60) and controls (n = 61). Information on the family history of haemophilia and inhibitors, FVIII gene mutations, prenatal and perinatal events (villocentesis, amniocentesis, prematurity and cesarean birth), breast-feeding and its duration, clinical and infusional history [age at treatment onset, number of EDs, mode of treatment delivery and regimen, infusions given during infections or vaccinations, surgical procedures and central nervous system (CNS) bleeding] was retrospectively collected from the medical records and completed during a prospective follow-up. Additional blood samples were obtained and patients, parents and caregivers were interviewed, if necessary. Severe molecular defects in the FVIII gene (large deletions, non-sense mutations and intron 22 inversions) were collectively referred to as null mutations (Oldenburg et al, 2004). FVIII infusions at doses ranging from 25 to 40 IU/kg regularly given two to three times per week or every other day were referred to as prophylaxis.
Continuous variables were expressed as medians and ranges and compared using the Mann–Whitney U-test. Categorical variables were expressed as frequency and percentage values and compared by the chi-square test or the Fischer exact test. Putative risk factors were considered to be: a family history of inhibitors, null mutations in the FVIII gene, amniocentesis or villocentesis, premature and caesarean birth, breast-feeding, age at FVIII treatment onset, surgery, FVIII infusions during infections or vaccinations and CNS bleeding. In order to evaluate prophylactic treatment as a putative risk factor, statistical analysis was also performed in a subgroup of cases (n = 25) who started prophylaxis prior to inhibitor development (n = 7) or were potentially eligible for prophylaxis because they were inhibitor-free up to the age of 35 months (n = 18). This age limit was chosen because it was the upper limit of the age range at the time of prophylaxis onset in cases and the median age at prophylaxis onset in controls. The duration of breast-feeding was chosen according to the median value observed in controls (6 months or less, and longer than 6 months) and the non-breast-fed group was used as reference. The strata of age at first FVIII infusion were chosen on the basis of the distribution tertiles in controls (older than 16, 16 to 11 and younger than 11 months) using the oldest group as the reference. FVIII treatment delivery by continuous infusion was not analysed as a putative risk factor because it was never used in the patients entered in this study. Unconditional logistic regression was used to calculate the relative risk of inhibitor development, expressed as crude odds ratios (ORs). The effect of each variable adjusted for the others was assessed by multivariate analysis and results were expressed as adjusted ORs and 95% confidence intervals (CI). P-values < 0·05 were considered significant.
The main characteristics of the cases are shown in Table I. Inhibitors were first detected at the median age of 23 months (range: 1–94) after a median of 16 EDs (range: 5–150). Overall, the historical inhibitor peaks ranged between 0·6 and 500 BU/ml (median: 17), and were <5 BU/ml in 10 cases (17%). Transient inhibitors were 7 (12%), all of which were <5 BU/ml. In the controls, inhibitors were never detected in 48 (79%) patients who exceeded 150 EDs and in 13 (21%) who did not. Since inhibitor development occurred in cases up to 150 EDs, the latter 13 controls were excluded from the statistical analysis. Controls were older than cases at study entry whereas severity of haemophilia was similar (Table I). Two patients, one case and one control, were siblings; two siblings were also in the control group.
|Cases (n = 60)||Controls (n = 48)||P-value|
|Age at study entry, months* (range)||77 (2–151)||106 (38–200)||0·007|
|Italian origin (%)||57 (95)||46 (96)||NS|
|Plasma FVIII <1% (%)||49 (82)||39 (81)||NS|
|Family history of haemophilia (%)||23 (38)||15 (31)||NS|
|Family history of inhibitors (%)||12 (20)||1 (2)||0·001|
|Null mutations† (%)||43/52 (83)||27/42 (64)||0·04|
|Amniocentesis or villocentesis (%)||7 (12)||9 (19)||NS|
|Premature birth (%)||7 (12)||3 (6)||NS|
|Caesarean birth (%)||25 (42)||19 (40)||NS|
|Breast-fed (%)||47 (78)||34 (71)||NS|
|Duration, months* (range)||4‡ (1–24)||6 (1–36)||NS|
|Age at first FVIII infusion, months* (range)||11 (2 days-64)||13 (1 day-57)||NS|
|FVIII infusions during infections or vaccinations (%)||12‡ (20)||11§ (23)||NS|
|Surgery (%)||15‡ (25)||11§ (23)||NS|
|At first FVIII infusion (%)||7 (12)||2 (4)||NS|
|CNS bleeding (%)||5‡ (8)||2§ (4)||NS|
|Prophylaxis (%)||7‡ (12)||34§ (71)||<0·0001|
A family history of inhibitors and null mutations in the FVIII gene were more frequent in cases than in controls (20% vs. 2%; P = 0·001 and 83% vs. 64%; P = 0·04, respectively, Table I). Surgery was the reason for FVIII treatment onset in seven cases (12%), all with high-responding inhibitors, but no statistical significance was found in comparison with controls (Table I). The prevalences of amniocentesis or villocentesis, premature or caesarean birth and breast-feeding were similar in cases and controls (Table I). No difference was also found for age at first FVIII infusion, duration of breast-feeding, frequency of FVIII infusions during infections or vaccinations and CNS bleeding prior to inhibitor development (Table I). By univariate analysis, family history of inhibitors and null mutations in the FVIII gene were associated with an increased risk of inhibitors, however, these associations were not confirmed by multivariate analysis (Table II). No association was found between duration of breast-feeding and the risk of inhibitor development (Table II). A trend was observed comparing cases and controls according to the distribution tertiles of age at first FVIII exposure in controls (older than 16 months: reference; 16–11 months: OR 1·7, 95% CI 0·6–4·7; 11 months or younger: OR 2·8, 95% CI 1·0–7·6). This inverse relationship between age at treatment onset and likelihood of inhibitor development did not reach statistical significance after adjusting for other variables (Table II).
|Cases (n = 60)||Controls (n = 48)||Crude OR (95% CI)||Adjusted OR* (95% CI)|
|Family history of inhibitors (%)||12 (20)||1 (2)||12·0 (1·5–96·0)||6·8 (0·8–60·0)|
|Null mutations (%)||43/52 (83)||27/42 (64)||2·6 (1·0–6·9)||2·2 (0·8–6·1)|
|No (%)||13 (22)||14 (29)||1 (ref.)||1 (ref.)|
|≤6 months (%)||29† (48)||20 (42)||1·6 (0·6–4·0)||1·6 (0·5–5·0)|
|>6 months (%)||18† (30)||14 (29)||1·4 (0·5–3·9)||1·9 (0·6–6·4)|
|Age at first FVIII infusion|
|>16 months (%)||11 (18)||16 (33)||1 (ref.)||1 (ref.)|
|11–16 months (%)||20 (33)||17 (36)||1·7 (0·6–4·7)||2·5 (0·7–8·9)|
|<11 months (%)||29 (49)||15 (31)||2·8 (1·0–7·6)||3·3 (0·9–12·0)|
Prophylaxis as putative risk factor
Prophylaxis was started within the first 150 EDs in 71% of controls whereas it was started prior to inhibitor development in seven cases (12%, P < 0·0001, Table I); three of seven had low-responding and two had transient inhibitors. Cases were younger at prophylaxis onset (median age: 15 months, range: 11–35) than controls (median age: 35 months, range: 5–132; P = 0·004). In order to evaluate whether or not the implementation of prophylaxis was a risk factor for inhibitor development, a statistical analysis was performed in 48 controls and in a subgroup of 25 cases selected because they received prophylactic treatment (n = 7) or were potentially eligible for prophylaxis because they were inhibitor-free up to the age of 35 months (n = 18). This age limit was chosen because it was the upper limit of the age range at prophylaxis onset in cases and the median age at prophylaxis onset in controls. The subgroup of 25 cases was similar to controls with respect to the age at study entry (median: 109 months, range: 19–151), the severity of haemophilia (FVIII<1%: 17, 68%) and the age at first FVIII infusion (median: 13 months, range: 5–64). The results of univariate and multivariate analysis are shown in Table III. The association between family history of inhibitors and increased risk of inhibitors was confirmed in this subgroup of patients by univariate analysis, while no association was found with null mutations in the FVIII gene and age at first FVIII infusion (Table III). Patients who started prophylaxis had a significantly smaller risk of developing inhibitors than patients treated on demand, and the risk remained significantly smaller after adjusting for other variables (Table III).
|Cases (n = 25§)||Controls (n = 48)||Crude OR (95% CI)||Adjusted OR* (95% CI)|
|Family history of inhibitors (%)||6 (24)||1 (2)||14·8 (1·7–131·7)||4·5 (0·3–62·8)|
|Null mutations (%)||15/18 (83)||27/42 (64)||2·8 (0·7–11·2)||2·5 (0·5–12·5)|
|No (%)||5 (20)||14 (29)||1 (ref.)||1 (ref.)|
|≤6 months (%)||12† (48)||20 (42)||1·7 (0·5–5·8)||2·2 (0·4–12·8)|
|>6 months (%)||8† (32)||14 (29)||1·6 (0·4–6·1)||4·0 (0·7–23·1)|
|Age at first FVIII infusion|
|>16 months (%)||10 (40)||16 (33)||1 (ref.)||1 (ref.)|
|11–16 months (%)||6 (24)||17 (36)||0·6 (0·2–1·9)||0·9 (0·2–4·7)|
|<11 months (%)||9 (36)||15 (31)||1·0 (0·3–3·0)||1·2 (0·2–6·9)|
|Prophylaxis (%)||7 (28)||34‡ (71)||0·2 (0·06–0·5)||0·2 (0·06–0·9)|
Finally, to evaluate the impact on inhibitor risk of early implementation of prophylaxis, the subgroup of 25 cases was compared with a subgroup of 32 controls, excluding those who started prophylaxis after 35 months of age (16/48). This subgroup of 32 controls was similar to the 25 cases for the age at study entry (median: 102 months, range: 38–184), the severity of haemophilia (FVIII<1%: 25, 78%) and the age at first FVIII infusion (median: 12 months, range: 2 days to 38 months). By univariate analysis, patients who started prophylaxis before the age of 35 months (7/25 cases, 28% and 18/32 controls, 56%) had a lower inhibitor risk than those treated on demand (OR 0·3, 95% CI: 0·1–0·9).
Many environmental factors may contribute to the onset of inhibitors, however it is difficult to assess their interactions with the genetic predisposition. On a national basis, we chose to evaluate this interaction by designing a case–control study that included a cohort of haemophiliacs who had uniform characteristics with respect to ethnicity, age, disease severity and type of FVIII used. Taking into account that antenatal and neonatal events can be precisely reported in paediatric patients only, we focused the study on children treated exclusively with rFVIII (Santagostino & Mannucci, 2000). Hence, product-related factors could not be evaluated in this study. We confirmed that the genetic background affects inhibitor development, a family history of inhibitors and the detection of null gene mutations being associated with an increased risk. At variance with the findings of large cross-sectional studies (Schwaab et al, 1995; Kemball-Cook et al, 1998), these predisposing factors were not statistically significant after adjustment for non-genetic factors, perhaps because a relatively small number of patients was included in the analysis. Among environmental factors, antenatal exposure to maternal FVIII and breast-feeding have been considered as putative candidates, surmising that induction of immune tolerance is facilitated in neonates who have an immature immune system (Yee & Lee, 2000). Human milk favours a normal gastrointestinal development and oral immune tolerance (Hanson, 1998) and contains a fat globule protein that has a strong sequence homology with FVIII (Larocca et al, 1991). No protective effect of breast-feeding on inhibitor formation was found, in agreement with the results of a previous cohort study in Swedish patients with severe haemophilia (Knobe et al, 2002). Invasive procedures during the antenatal/perinatal period and FVIII infusions during vaccinations or infections were evaluated for the first time in our series, showing no role on inhibitor development. Since catheter-related infections were included in the analysis, our results indirectly suggest that infections associated with the use of central venous catheters in haemophilic children (Valentino et al, 2004) are not associated with a higher inhibitor risk. FVIII treatment for severe bleeding or surgery has been suspected to influence inhibitor formation because extensive tissue damage and inflammation may trigger an antibody response against extravascular FVIII (Oldenburg et al, 2004). The inhibitor risk was not significantly affected by these conditions.
Previous studies carried out in two cohorts of Spanish and Dutch haemophiliacs (Lorenzo et al, 2001; van der Bom et al, 2003) showed an inverse relationship between the age at treatment onset and the likelihood of inhibitor development. A similar trend was found in this study comparing cases and controls on the basis of the distribution tertiles of age at first FVIII exposure in controls (>16, 16–11 and <11 months), however differences were not statistically significant after adjusting for genetic factors. Both of the previous studies arbitrarily defined the strata of age at first FVIII exposure (>12, 12–7, ≤6 months), however, no significant trend was observed in this study by computing age at first FVIII exposure according to the same age strata (older than 12 months: reference; 12–7 months: OR 1·1, 95% CI 0·5–2·6; 6 months or younger: OR 2·2, 95% CI 0·8–6·2). Apart from the different study design, the main difference between this and previous studies is represented by the type of FVIII products used at the time of treatment onset: rFVIII in our series, low/intermediate-purity concentrates in the oldest patients and high-purity or recombinant products in the youngest patients of the two aforementioned series (Lorenzo et al, 2001; van der Bom et al, 2003). Furthermore, such confounders as the type of FVIII gene mutations and other environmental conditions were not taken into account in the previous studies (Lorenzo et al, 2001; van der Bom et al, 2003).
Early prophylaxis is recommended for severe haemophiliacs to prevent joint damage and progression towards arthropathy (Santagostino & Mannucci, 2000; National Hemophilia Foundation, 2001), rendering it difficult to design controlled prospective studies comparing the inhibitor risk in children on prophylactic versus on-demand treatment. This case–control study enabled us to evaluate whether or not the implementation of prophylactic treatment plays a role on inhibitor formation. As expected, a relevant proportion of controls were on prophylaxis, whereas inhibitor development prevented the start of prophylaxis in the majority of cases. Considering that prophylaxis was started prior to inhibitor development in seven cases only, all aged 35 months or younger, and that 35 months was the median age at prophylaxis onset in controls, we analysed a subgroup including only those cases who were on prophylaxis or were potentially eligible for prophylaxis because they were free from inhibitors up to the age of 35 months. Prophylaxis gave a protective effect on inhibitor development that remained statistically significant after adjustment for genetic and environmental factors, including FVIII gene mutations and age at first FVIII exposure. Interestingly, three of the seven cases who started prophylaxis developed low-response antibodies, suggesting that perhaps prophylactic treatment favours this pattern of inhibitor behaviour. Our results in the subgroup of patients who started prophylaxis before 3 years of age hinted at a favourable impact of early prophylaxis on inhibitor risk that should be investigated in larger series.
On the basis of their findings of an inverse relationship between age at first treatment and inhibitor development, Lorenzo et al (2001) and van der Bom et al (2003) considered the option of delaying the exposure to FVIII or using non-FVIII containing products. Following these considerations, recombinant activated FVII was used by Rivard et al (2004) in 11 infants in order to postpone FVIII therapy, however six required FVIII treatment to control bleeding and, subsequently, four developed inhibitors. In our opinion, these approaches of delayed or alternative treatments might expose haemophilic children to complications and long-term sequelae, because FVIII products represent the most effective option for treatment and prevention of bleeding; furthermore, inhibitor development seems to be deferred but not prevented by this strategy (Rivard et al, 2004). On the other hand, the protective effect on inhibitor formation shown by prophylaxis in this study seems to be an additional advantage to prompt the use of this treatment regimen in children with severe haemophilia.
The authors thank Dr Matteo Luciani (‘Bambino Gesù’ Children Hospital, Rome, Italy) and Dr Elio Boeri (‘G. Gaslini’ Hospital, Genoa, Italy) for their contribution to the clinical data collection; Dr Paolo Bucciarelli (Maggiore Hospital of Milan, Milan, Italy) who took part in the statistical analysis; Prof. Maurizio Margaglione (Department of Biological Science, University of Foggia, Italy), Prof. Giuseppe Castaldo (CEINGE – Biochemical Department, University of Naples, Italy) and Prof. Flora Peyvandi (University of Milan, Milan, Italy) who performed FVIII gene molecular analysis for this study.