Advances in the understanding of acquired haemophilia A: implications for clinical practice

Authors


Peter W. Collins, Arthur Bloom Haemophilia Centre, University Hospital of Wales and School of Medicine, Cardiff University, Heath Park, Cardiff, UK. E-mail: peter.Collins@cardiffandvale.wales.nhs.uk

Summary

Acquired haemophilia A is an auto-immune disease caused by an inhibitory antibody to factor VIII. Patients with an acquired factor VIII inhibitor are at risk of life- and limb-threatening bleeding until the inhibitor has been eradicated. Management relies on rapid and accurate diagnosis, control of bleeding episodes, investigation for a precipitating cause and eradication of the inhibitor by immunosuppression. Patients should always be managed jointly with a specialist centre even if they present without overt bleeding. Despite an extensive literature, few controlled data are available and management guidelines are predominantly based on case reports, retrospective cohorts and expert opinion. This paper reviews the current literature on incidence, pathogenesis, diagnosis, haemostatic therapy and inhibitor eradication strategies. Potential future developments are discussed.

Acquired haemophilia A (AHA) is an autoimmune disease caused by an autoantibody to factor VIII (FVIII). It is associated with high morbidity and mortality secondary to bleeding, the age of the patient, underlying diseases and the toxic effects of immunosuppression (Green & Lechner, 1981; Morrison & Ludlam, 1995; Hay, 1998; Delgado et al, 2003; Kessler & Asatiani, 2006; Collins et al, 2007; Franchini & Lippi, 2008; Huth-Kuhne et al, 2009). Early recognition, rapid diagnosis and prompt referral to a specialist centre are important to facilitate the optimum treatment of bleeds and improve outcomes. Education of specialists, such as obstetricians, rheumatologists and care of the elderly physicians, about AHA is an important role for haematologists.

Incidence

The incidence of AHA has been reported as 1·48 (Collins et al, 2007) and 1·34 (Collins et al, 2004) per million/year in the only two studies in which patients were linked to a defined population. Both these studies are from the UK and incidence data from other populations are awaited. The incidence increases with age and is estimated to be 0·045/million per year in children under 16 compared to14·7/million per year in people over 85 years (Collins et al, 2007). It is likely that AHA is under diagnosed, especially in elderly patients. The incidence in males and females is similar, except in 20–40 year olds where the effect of pregnancy results in a preponderance of females (Green & Lechner, 1981; Collins et al, 2007) (Fig 1). Age contributes to the relatively poor prognosis in this patient group (Delgado et al, 2003).

Figure 1.

 Age related incidence of acquired haemophilia A. Data are shown for the percentage of patients presenting with acquired haemophilia A in each decade of life in two large cohorts, a treatment study and a combined analysis of 20 cohorts. Black bars (Green & Lechner, 1981), white bars (Collins et al, 2007), lined bars (Morrison et al, 1993), hatched bars (Delgado et al, 2003) and grey boxes interim data from EACH2 registry (periods 0–20, 20–40 and 40–60 are combined) (Levesque et al, 2009).

Pathogenesis

Acquired haemophilia A is caused by polyclonal auto-antibodies to FVIII, usually IgG 1 and 4, directed against the A2 and or C2 domains (Scandella et al, 1989; Matsumoto et al, 2001; Reding et al, 2002). However, antibodies to FVIII have been demonstrated in 17% of healthy people (Algiman et al, 1992) and hence not all antibodies are pathogenic (Scandella et al, 1989). Some auto-antibodies to FVIII have been shown to mediate FVIII proteolysis (Wootla et al, 2008). The disorder is associated with other auto-immune diseases and older age, suggesting a more generalised immune dysregulation. The mechanisms that result in some people developing or failing to regulate pre-existing anti-FVIII auto-antibodies is not fully understood although variation in the F8 gene has been reported (Tiede et al, 2004). Studies have also demonstrated an increased frequency of cytotoxic T lymphocyte antigen (CTLA)-4 49 A/G, a molecule that plays an important role in T cell activation, in patients with AHA compared to controls (Pavlova et al, 2008) and better understanding of the immune mechanisms involved are likely to improve treatment.

Inhibition usually follows complex kinetics and residual FVIII can often be measured; in contrast, inhibitors in congenital haemophilia usually follows first order kinetics (Gawryl & Hoyer, 1982; Scandella et al, 1989). This has implications for measuring the titre of anti-FVIII antibodies in patients with AHA and may play a role in the typical bleeding pattern of AHA.

Diagnosis

Acquired haemophilia A is usually diagnosed following investigation of abnormal bleeding, although 7% of patients presented with an abnormal routine coagulation test rather than bleeding in an interim analysis of the European Acquired Haemophilia Registry (EACH2, http://www.each2registry.org) (Levesque et al, 2009). The diagnosis must be made promptly to minimise the time a patient is at risk of bleeding and to avoid non-essential invasive procedures. Interim data from EACH2 reported a median (5th–95th percentile) delay of 3 (0–58) d between onset of bleeding and diagnosis and 1 (0–69) d between the first abnormal activated partial thromboplastin time (aPTT) and diagnosis (Levesque et al, 2009), suggesting significant delay in a proportion of patients.

Clinical features

Patients typically present with subcutaneous bruising, mucosal and deep soft tissue bleeding, such as intracranial haemorrhage and muscle and retroperitoneal haematoma (Fig 2). Bleeding following invasive procedures is almost inevitable in undiagnosed patients but haemarthoses are relatively uncommon (Green & Lechner, 1981; Morrison et al, 1993; Morrison & Ludlam, 1995; Hay, 1998; Delgado et al, 2003; Kessler & Asatiani, 2006) (Fig 3). Occasionally, patients are found with no bleeding manifestations (Collins et al, 2007). The severity of bleeding is unpredictable and, although between 25% and 33% of patients do not require any haemostatic therapy (Lottenberg et al, 1987; Collins et al, 2004, 2007), they remain at risk of life threatening bleeding until the inhibitor is eradicated. The FVIII level and inhibitor titres are poor predictors of bleeding risk (Collins et al, 2007).

Figure 2.

 Bleeding symptoms at presentation in acquired haemophilia. The bleeding symptoms seen at presentation in a consecutive cohort of 172 patients with acquired haemophilia A (Collins et al, 2007). Subcut, subcutaneous bleed; GI tract, gastrointestinal bleeding; GU tract, genito-urinary bleeding; none, bleed did not require haemostatic therapy; fatal, bleeding that was the main cause of death; ICH, intracranial haemorrhage.

Figure 3.

 Investigation for possible acquired haemophilia A. The interpretation of coagulation test results is shown. Failure of the aPTT to correct is typical of acquired haemophilia but a full correction, even on an incubated sample, does not exclude the diagnosis and further testing of factor VIII and inhibitor levels should be undertaken if the clinical situation is suggestive of this disorder.

Fatal bleeds, including intracranial, deep soft tissue, mucosal (gastrointestinal, lung or urogenital) or following invasive procedures, were reported in 22% of cases in older studies but more recently this had decreased to 8%, probably because of increased availability of efficacious haemostatic agents and possibly an improved awareness of the disease (Green & Lechner, 1981; Lottenberg et al, 1987; Collins et al, 2007). Fatal bleeds may occur at any time after presentation until the inhibitor has been eradicated (Collins et al, 2007). Bleeding can also lead to compartment syndrome and neurovascular damage (Lottenberg et al, 1987).

There have been a number of reports of anticoagulant and anti-platelet drugs either masking or being associated with AHA (Uggla et al, 2003; Dragani et al, 2004; Haj et al, 2004; Vadikolia et al, 2007). Patients who present with bruising or bleeding whilst taking warfarin or an anti-platelet agent should have an aPTT performed in addition to an International Normalized Ratio (INR) and, if inappropriately prolonged, investigated further with mixing studies and factor levels.

Laboratory diagnosis

Laboratory investigation typically reveals a prolonged aPTT with a normal prothrombin time. Anti-FVIII antibodies are time- and temperature-dependent and the aPTT does not correct with normal plasma, especially after incubation. In some cases correction of the aPTT does occur and this finding alone cannot be used to exclude AHA. The diagnosis is confirmed by the finding of a low FVIII and a raised inhibitor titre on Bethesda assay (Fig 3). Due to the complex enzyme kinetics, residual FVIII can often be measured and establishing an accurate inhibitor titre can be difficult. It is routine practice to report the titre based on the dilution in the Bethesda assay that is closest to 50% inhibition (Hay et al, 2000; Huth-Kuhne et al, 2009). The FVIII antibody may interfere with the measurement of other intrinsic factors. Dilution experiments will demonstrate a progressive increase in these apparently decreased coagulation factors whilst FVIII remains low (Morrison & Ludlam, 1995; Hay, 1998; Kazmi et al, 1998; Delgado et al, 2003; Kessler & Asatiani, 2006). A lupus anticoagulant may also interfere with coagulation factor assays potentially leading to diagnostic difficulties (Kazmi et al, 1998). An enzyme-linked immunosorbent assay may be useful in complicated cases (Sahud et al, 2007).

Clinical associations

Acquired haemophilia A is associated with autoimmune diseases (such as rheumatoid arthritis, polymyalgia rheumatica and systemic lupus erythematosis), malignancy (occasionally occult), pregnancy and dermatological disorders, such as pemphigoid. The reported association with commonly used drugs, such as penicillin, is difficult to assess and may be due to chance associations. In about half of cases no underlying cause is found (Bossi et al, 1998; Delgado et al, 2003; Green & Lechner, 1981; Hauser et al, 1995; Hay, 1998; Italian Association of Haemophilia Centres (AICE) (2006); Kessler & Asatiani, 2006; Morrison et al, 1993; Sallah & Wan, 2001; Collins et al, 2007), especially in elderly patients (Collins et al, 2007), possibly because they are less intensively investigated, or that old age is an independent risk factor.

Pregnancy-related acquired haemophilia

Acquired haemophilia A is a very rare complication of pregnancy, estimated to affect one in 350 000 births in the UK (Collins et al, 2007) and 20 cases in 15 years were reported in a survey of 42 specialist Italian centres (AICE, 2006). Patients usually present with bleeding at the time of delivery or within the first 1–4 months postpartum, although some present ante-partum and some up to a year postdelivery (Hauser et al, 1995; Michiels et al, 1997; Solymoss, 1998; AICE 2006). The reasons for the late presentations are unclear, however, this may represent delayed diagnosis rather than late onset of AHA in some patients.

Some reports suggest that AHA in pregnancy has a different natural history and response to inhibitor eradication therapy (Hauser et al, 1995). Retrospective reviews have reported that pregnancy-related AHA takes longer to achieve remission compared to other underlying aetiologies, although conversely, spontaneous remissions are recognised (Hauser et al, 1995; Solymoss, 1998; AICE, 2006). Similar to other patients with AHA, the available data are insufficient to convincingly conclude that outcome is affected by choice of immunosuppressive agents, and treatment decisions should take into account the age of the patients and the potential side effects of drugs in women of child bearing age (Hay et al, 2006). Rituximab has been used successfully in postpartum AHA (Dedeken et al, 2009).

Relapse in subsequent pregnancies appears to be relatively uncommon but women should be warned that this is a possibility. In one study it was observed that AHA recurred in four of six subsequent pregnancies in three patients (Solymoss, 1998), however no relapses were reported in nine subsequent pregnancies in another study (Coller et al, 1981) and an Italian Registry reported no relapses amongst four patients (AICE, 2006). The antibody may affect the FVIII level of the fetus and this must be considered at the time of delivery (Ries et al, 1995; Lulla et al, 2009).

Treatment of bleeds should follow the principles outlined for AHA in general (Hauser et al, 1995; Solymoss, 1998; AICE, 2006) but caution about the risk of venous thromboembolism associated with bypassing agents in the postpartum period should be borne in mind.

Acquired haemophilia in children

Acquired haemophilia A in children is very uncommon, estimated at 0·045/million per year (Collins et al, 2007). A survey of haemophilia centres in USA reported six cases and a literature review revealed another eight presumed or definite cases (Moraca & Ragni, 2002). A large retrospective study reported a further six cases (Green & Lechner, 1981). Patients appeared to present with a bleeding pattern typical for AHA and response to immunosuppression appears to be similar to adults.

Treatment

The main principles of treatment for AHA are to control bleeding, eradicate the inhibitor, treat underlying disorders and protect the patient against trauma and non-essential invasive procedures. Patients should be treated in collaboration with a centre experienced in the management of inhibitors even if the initial presentation appears to be benign. Patient education is important so that symptoms are recognised and reported early (Hay et al, 2000, 2006; Franchini & Lippi, 2008; Huth-Kuhne et al, 2009).

Haemostatic management

Bleeding episodes may be very severe and prompt haemostatic control is vital to reduce morbidity and mortality. Treatment of bleeds in AHA may be difficult and should be supervised by a specialist in the field. Available haemostatic agents do not have predictable efficacy, hence regular clinical review supported by appropriate imaging and measurement of haemoglobin level is crucial for optimal outcomes (Huth-Kuhne et al, 2009).

Bleed management depends on site and severity. Often, therapy needs to be continued at a reduced dose after initial haemostasis has been achieved to prevent recurrence, especially after intracranial, muscle and retroperitoneal bleeds. Local measures to control bleeds should be used. Mucosal haemorrhage will benefit from concomitant therapy with an anti-fibrinolytic agent and topical fibrin glue may be useful in some cases.

The two options for haemostatic control are the use of bypassing agents and strategies to raise the level of circulating FVIII (Hay et al, 2006).

Bypassing agents.  Bypassing agents are currently the most commonly used first-line treatment and both recombinant activated factor VII (rFVIIa; Novoseven Novo Nordisk A/S, Bagsvaerd, Denmark) and FVIII inhibitor bypassing activity (FEIBA) (the only currently available activated prothrombin complex concentrate) have been shown to be efficacious in AHA (Hay et al, 1997; Sallah, 2004; Tjonnfjord, 2004; Sumner et al, 2007). A recent retrospective analysis of 139 patients treated with rFVIIa has been published (Sumner et al, 2007) and included cases from the compassionate used programme (n = 61) (Hay et al, 1997), the Hemophilia and Thrombosis Research Society (HTRS) registry (n = 9) and the published literature (n = 69). There were 182 bleeding episodes where it was possible to assess the efficacy of rFVIIa. In the 103 episodes where rFVIIa was used as first-line therapy, it was effective or partially effective in 95%. When used as second-line therapy, an effective or partially effective response was described in 80% of cases. In 57 surgeries an effective or partially effective response was reported in 86% of cases. For the 61 patients in whom data were available, the mean duration of treatment was 6 (range 1–33) d (Sumner et al, 2007). Retrospective studies with FEIBA describe 34 severe and moderate bleeds treated, in the main, with 75 u/kg 8–12 hourly. A median of six infusions were needed for moderate bleeds with 100% haemostatic efficacy at a median of 36 h compared to 10 infusions for severe bleeds with 76% haemostatic control at a median of 48 h (Sallah, 2004).

The reports on the efficacy of rFVIIa and FEIBA cannot be compared and there are no data to suggest that either agent has a superior haemostatic efficacy. Both bypassing agents have proven efficacy in AHA but neither agent has predictable efficacy in all cases and close clinical monitoring of the patient is essential. The choice of agent should depend on considerations such as previous patient responses, dosing schedule, use of plasma-derived products and cost. If first-line therapy fails the alternative bypassing agent may be successful and should be tried at a relatively early stage.

Both rFVIIa and FEIBA are associated with thrombotic events (Aledort, 2004, 2005). A recent analysis of 139 AHA patients treated with rFVIIa reported 12 thrombotic events, predominantly arterial, in 10 patients, four of whom died (Sumner et al, 2007). A 10-year compilation of thrombotic events associated with FEIBA reported one episode of disseminated intravascular coagulation, one myocardial infarction and one venous thrombosis (Ehrlich et al, 2002). It is not possible to compare the rate of thrombosis with that observed in association with rFVIIa because the number of patients treated is not known. The risk of thrombosis in patients with AHA treated with bypassing agents appears to be significantly higher than in congenital haemophilia, probably because of the additional risk factors associated with elderly patients and the complex clinical situation of many patients with AHA.

A strategy to increase doses of rFVIIa up to 270 μg/kg has been described successfully in the management of congenital haemophilia with inhibitors (Parameswaran et al, 2005;Santagostino et al, 2006). This approach should be used very cautiously in patients with AHA and has not been shown to be safe in this patient group. In the management of bleeds uncontrolled by conventional doses, escalation may be justifiable on a case-by-case basis.

An important limitation of bypassing agents is that there is no currently validated laboratory monitoring technique. The use of thrombin generation assays (Turecek et al, 2003; Varadi et al, 2003) and modified thromboelastographic assays (Sorensen & Ingerslev, 2004, 2005) hold promise. Laboratory changes in assay parameters can be demonstrated with ex vivo spiking and plasma taken from patients infused with bypassing agents but no data have been published that convincingly ties these results to haemostatic efficacy in AHA. Some preliminary data suggest that thromboelastography is sensitive to rFVIIa but does not always predict clinical response (Dehmel et al, 2008). Further data linking these assays to clinical endpoints are awaited. Extrapolation of results from inhibitors in congenital haemophilia to AHA may not be valid due to the different inhibitor kinetics and bleeding phenotypes. If, however, it is shown that improvement of these assays with bypassing agents correlates with clinical efficacy a significant advance in the management of patients with FVIII inhibitors will have been made.

Human FVIII.  Human FVIII will usually be inadequate haemostatic therapy unless the inhibitor titre is low. The dose of FVIII required will need to be sufficient to overcome the inhibitor and provide an adequate haemostatic level. Although formulae have been suggested for calculating the dose (Kessler & Asatiani, 2006) the inaccuracies inherent in the laboratory measurement of inhibitor titres in AHA makes these, at best, very rough approximations and regular monitoring of plasma FVIII level and clinical response is required.

The use of human FVIII in combination with immunoabsorption is more likely to result in haemostatic FVIII levels despite higher anti-FVIII inhibitor titres. This treatment strategy may be useful as first-line or if bypassing agents have failed, although it is available in only a very limited number of centres. The technique relies on adequate venous access and FVIII levels must be closely monitored (Guillet et al, 2001; Freedman et al, 2003; Rivard et al, 2003; Zeitler et al, 2006).

Porcine FVIII.  In AHA, the inhibitor titre to porcine FVIII is usually 5–10% of the human titre and this often means that porcine FVIII can achieve haemostatic levels in situations where human FVIII is ineffective (Morrison et al, 1993; Hay et al, 1996). Porcine FVIII has been shown to have excellent or good haemostatic efficacy in 78% of 74 bleeds but no response in 9% (Morrison et al, 1993). It has been used successfully as a continuous infusion (O’Gorman et al, 2001). This agent, however, is no longer available. A recombinant B-domain deleted porcine FVIII is under investigation and trials in AHA are awaited (Mahlangu et al, 2007).

Desmopressin.  Some patients with a low titre inhibitor and measurable baseline FVIII may respond to a desmopressin (DDAVP) infusion. A literature review reporting on 22 cases found that, for 11 patients with FVIII >5 iu/dl, DDAVP resulted in an increased FVIII level to between 15 and 140 iu/dl. Five patients with a Bethesda titre of <2 Bethesda Units (BU) responded best with peak FVIII levels >80 iu/dl and a half-life of 4–6 h. In the 22 patients no clinical response was reported in seven, all whom had a FVIII level <iu/dl and a rise following DDAVP to between 0 and 6 iu/dl. In contrast, three other patients with baseline FVIII levels between 1 and 2·9 iu/dl had increases in FVIII to between 15 and 27 iu/dl and good clinical efficacy. The response to DDAVP is clearly unpredictable but its use can be considered for minor bleeding episodes (Mudad & Kane, 1993).

Management of surgery.  Invasive procedures are associated with significant risk and haemostasis cannot be guaranteed. Only essential procedures should be considered and, even then, the benefits carefully weighed against the risks. Haemostatic options for surgery include the use of bypassing agents (Hay et al, 1997; Tjonnfjord, 2004), immunoabsorption with FVIII infusion and previously porcine FVIII (Morrison et al, 1993).

Inhibitor eradication

Immunosuppressive therapy to eradicate the inhibitor in AHA should be undertaken as soon as the diagnosis has been made (Hay et al, 2006; Franchini & Lippi, 2008; Huth-Kuhne et al, 2009). Whilst numerous publications can be found in the literature, the data are often difficult to interpret because different endpoints and definitions are used and studies are predominantly reports of cohorts without controls. The majority of papers are case reports, single centre cohort studies or retrospective surveys (Table I) (Hultin et al, 1976; Green & Lechner, 1981; Spero et al, 1981; Hart et al, 1988; Lian et al, 1989, 2002; Schwartz et al, 1995; Schulman et al, 1996; Shaffer & Phillips, 1997; Sohngen et al, 1997; Bossi et al, 1998; Green, 1998; Sallah et al, 1998; Bayer et al, 1999; Saxena et al, 2000; Yee et al, 2000; Burnet et al, 2001; Dykes et al, 2001; Godreuil et al, 2001; Grunewald et al, 2001; Guillet et al, 2001; Sallah & Wan, 2001; Kain et al, 2002; Wiestner et al, 2002; Au et al, 2004; Collins et al, 2004, 2007, 2009; Huang et al, 2004; Stasi et al, 2004; Holme et al, 2005; Aggarwal et al, 2006; Di Bona et al, 2006; Ng et al, 2006; Pardos-Gea et al, 2006; Zeitler et al, 2006; Lak et al, 2009). This means that most reported cases are from specialist centres, potentially leading to the literature reflecting more severely affected patients. Furthermore, good outcomes are more likely to be reported by centres (and accepted by journals) than average or poor outcomes. The literature must, therefore, be treated with caution and the conclusions that can be drawn from many studies are limited.

Table I.   Literature review of inhibitor eradication therapy.
StudySteroidsSteroids and cytotoxicsStudy reference
CR/N% CRCR/N% CR
  1. Papers reporting on inhibitor eradication therapy identified from the literature are presented. Although the percentage of patients achieving complete remission does not take into account the length of time patients were followed, a broad comparison between treatment groups can be made. Not all patients from each study were included because treatment details or outcome could not be extracted. Interim results of the European Acquired Haemophilia Registry (EACH2) were not included in the aggregate because the results are likely to change when the final analysis is made, but are included for comparison because this is the largest cohort described to date.

 17/8 88Ji et al (2006)
 23/4 75Mazzucconi et al (2001)
 34/6 66Dykes et al (2001)
 47/16 44Spero et al (1981)
 57/10 70Green (1998)
 64/6 66Burnet et al (2001)
 71/3 33Sallah et al (1998)
 811/12 92Lian et al (1989)
 95/11 45Holme et al (2005)
109/9100Shaffer and Phillips (1997)
118/8100Bayer et al (1999)
126/6100Lian et al (2002)
134/5 80Huang et al (2004)
148/12 75Ng et al (2006)
155/8 631/3 33Di Bona et al (2006)
161/3 336/6100Grunewald et al (2001)
178/9 898/14 57Sallah and Wan (2001)
183/4 759/11 82Yee et al (2000)
193/31004/4100Saxena et al (2000)
209/10 9014/16 88Bossi et al (1998)
210/1  07/7100Sohngen et al (1997)
222/3 660/2  0Godreuil et al (2001)
232/3 6610/12 83Collins et al (2004)
249/12 7515/19 79Lak et al (2009)
2526/34 7635/45 78Collins et al (2007)
Aggregate of all studies96/134 72165/211 78 
Studies reporting both treatment arms68/90 76109/139 78 
EACH2 registry48/81 6047/57 82Collins et al (2009)

The main options for immunosuppression are steroids, cytotoxics (cyclophosphamide, azathioprine or combination therapy), rituximab, ciclosporin A, plasmapheresis or immunoabsorption and FVIII immune tolerance. These treatments have been combined in numerous ways. A regimen may be considered superior if more patients achieve complete remission (CR) or this is achieved more rapidly. Studies must be interpreted in the light of the finding that about 25% of patients achieved a spontaneous remission (Lottenberg et al, 1987), although there was significant morbidity and mortality in the untreated group, re-enforcing the recommendation that all patients should be treated as soon as the diagnosis is made to reduce the time patients are at risk of bleeding. Some patients require long term immunosuppression to prevent relapse.

Steroids and cytotoxic agents

The only prospective randomised study performed enrolled 31 patients who were treated with prednisolone 1 mg/kg for 3 weeks after which 10 patients were in CR (Green et al, 1993). Four patients were randomised to continuing treatment with prednisolone alone and this led to CR in three (75%). Of the 10 patients randomised to adding cyclophosphamide five (50%) achieved CR and of those in whom cyclophosphamide was substituted for prednisolone three out six (50%) achieved CR. There was no difference between the treatment arms and, therefore, no evidence to suggest that adding or changing to cyclophosphamide after 3 weeks was better than continuing with steroids alone. However, the study had insufficient power to demonstrate a difference unless it was very large (Green et al, 1993).

A non-randomised, retrospective national consecutive cohort study compared patients treated with steroids versus steroids and cytotoxics. The design of this study makes it less prone to selection bias. The 34 patients treated with steroids had 76% CR at a median (95% confidence interval, CI) of 49 (31–62) d compared to 78% CR at 39 (34–57) d for the steroids and cytotoxics group. There was no statistically significant difference between the treatment arms and mortality was not different (Collins et al, 2007).

A review that combined data from 20 publications reported that the use of steroids and cyclophosphamide resulted in more patients achieving CR compared to steroids alone (Delgado et al, 2003). The higher CR rate was not translated into a lower mortality. The authors speculated that this was due to increased toxicity of cyclophosphamide (Delgado et al, 2003). Interim data from the EACH2 registry also suggest that more patients achieve CR with a combination of steroids and cyclosphosphamide (82%) than with steroids alone (60%), although the median time to a negative inhibitor titre and normal FVIII in those that responded was virtually identical (33 and 34 d respectively) (Collins et al, 2009). These data should be treated with caution until the final analysis is performed and survival in this cohort has not yet been analysed.

The aggregated data available from uncontrolled cohorts published to date suggest similar results for combined steroids and cytotoxic agents: CR rate of 78% compared to 72% for steroids alone. If only studies that report the use of both treatments are included the response rates are almost indistinguishable; 78% for steroids and cytotoxic agents and 76% for steroids alone (Table I), the interim data from EACH2 have not been included. Regimens involving combination chemotherapy have been reported to have high success rates (Sohngen et al, 1997; Bayer et al, 1999; Saxena et al, 2000; Grunewald et al, 2001; Lian et al, 2002), but without comparative treatment groups the results must be treated with caution. Whichever regimen is used, 3 weeks appears to be too short a time to assess outcome because the median time to remission has been reproducibly been shown to be about 5 weeks (Collins et al, 2007, 2009).

Intravenous immunoglobulin

Intravenous immunoglobulin (IVIG) has been suggested to be a useful agent in AHA. A study on 16 assessable consecutive patients showed that three achieved an undetectable inhibitor titre and normal FVIII level (Schwartz et al, 1995). The starting inhibitor titre in these patients was 0·9, 1·0 and 1·0 BU/ml and one patient received concomitant steroids. In three further patients a fall in inhibitor titre was seen, although in only one did the FVIII level increase (Schwartz et al, 1995). A study of six patients treated with steroids and IVIG reported a CR rate of 66% (Dykes et al, 2001) similar to other reports of steroids alone (Table I). A larger study that compared non-randomised patients who either did or did not receive IVIG (Collins et al, 2007) and a literature review both showed no benefit of IVIG (Delgado et al, 2003). The available evidence suggests that IVIG as a single agent or in combination with steroids and cytotoxics is not useful in inhibitor eradication in AHA, although it possibly has a role in patients with very low inhibitor titres.

Rituximab

Rituximab has been used to treat cohorts of patients with AHA (Franchini & Lippi, 2008; Garvey, 2008). In one study three patients treated with rituximab and either steroids or steroids plus a cytotoxic agent achieved CR (Wiestner et al, 2002). Another study reported on 10 patients, of whom eight achieved CR and the two non-remitters responded to subsequent intravenous cyclophosphamide (Stasi et al, 2004). The response rate of 80% is very similar to other immunosuppressive therapies. A study in six patients treated with rituximab and steroids with or without cytotoxic agents found a CR rate of 100%. CR occurred at 1, 2, 4, 8, 36 and 52 weeks, times similar to previous reported results with other immunosuppressive agents (Adedayo et al, 2006). A literature review of 71 patients treated with rituximab and a variety of immunosuppressive agents found a response rate of over 90%, but the authors were cautious about interpreting the results and suggested that rituximab should be used as a second line agent in combination with steroids (Franchini & Lippi, 2008).

It has been suggested that rituximab may lead to more rapid remission than other therapies but without comparative patients this is difficult to clarify. An interim analysis of the EACH2 registry does not support this assertion. In 19 patients treated with first line rituximab (the largest cohort reported to date) and a variety of other immunosuppressive agents the CR rate was 63% and the median (interquartile range) time to negative inhibitor was 33 (25–51) d (Collins et al, 2009), a response indistinguishable from other immunosuppressive regimens.

The current data on rituximab are very difficult to interpret because it is almost always used in association with other immunosuppressive agents known to be efficacious in the treatment of AHA. There is no published convincing evidence that rituximab-based regimens result in either more patients achieving CR or a more rapid response.

Rituximab can be used as first- or second-line therapy and patients resistant to standard first-line regimens may respond. The toxicity profile of rituximab may make it preferable for some patients. There is no published evidence to support its use in patients with higher titre inhibitors as suggested by some authors (Aggarwal et al, 2006).

Ciclosporin A

A number of cases have been reported in which ciclopsorin A has induced CR following failed first line therapy (Hart et al, 1988; Pfliegler et al, 1989; Schulman et al, 1996; Au et al, 2004; Pardos-Gea et al, 2006).

Immune tolerance

The use of FVIII in conjunction with immunosuppressive agents in AHA is reported. A stated rationale for this approach is that FVIII may stimulate antibody-producing cells into division making them more susceptible to cytotoxic agents (Lian et al, 1989). The lack of adequate controls means that direct assessment of the role of FVIII can not be made.

A report of patients treated with three weekly infusions of FVIII combined with vincristine, cyclophosphamide and steroids resulted in a 92% CR rate in 12 patients after 1–3 courses (Lian et al, 1989). The same group, however, later published a report in six patients who were treated with vincristine, cyclophosphamide and steroids without FVIII and found 83% CR after 1–7 courses (Lian et al, 2002). These data are difficult to interpret, the CR rates are not dissimilar to other published studies given the numbers of patients involved but the time to remission appears to be relatively short. The role of FVIII in these results is unclear because the intensity of immunosuppression was greater than for many other published protocols.

Infusion of FVIII on a daily basis (30 iu/kg/d for 1 week, 20 iu/kg/d for a second week and 15 iu/kg/d for a third week) combined with intravenous cyclophosphamide and methylprednisolone reported CR in 93% of 14 patients after a median 4·6 weeks, compared to 67% remission at a median of 28·3 weeks in six historical controls treated with steroids ± cyclophosphamide (Nemes & Pitlik, 2000). Although this is a relatively high CR rate the median time to response is similar to studies that did not use FVIII and the median time of 28·3 weeks to CR in the controls appears to be long.

Taken together these reports are insufficient to conclude that immune tolerance with FVIII is beneficial in AHA and the cost of FVIII therapy in these protocols should be taken into account. Controlled studies appear to be the only way that this question can be answered.

Immunoabsorption

Immunoabsorption has been used to treat bleeding episodes by reducing the inhibitor titre but also as part of regimens aimed at inhibitor eradication. A cohort of 35 patients with AHA and severe bleeding was treated with a combination of oral cyclophosphamide 1–2 mg/kg daily, prednisolone 1 mg/kg daily, immunoadsorption days 1–5 weekly, IVIG 0·3 g/kg days 5–7 weekly and FVIII 100 iu/kg/d. Rapid control of bleeding was reported with an undetectable inhibitor at a median of 3 d (95% CI 2–4) and CR in 88% of patients at a median of 14 d (95% CI 12–17) (Zeitler et al, 2006). Although no control patients were included, the CR rate was similar to other protocols and the cost of the FVIII considerable, this treatment appeared to rapidly control bleeding and induce CR. It should be considered in severely bleeding patients, especially those unresponsive to bypassing agents, although the potential side effects of this treatment should be taken into account.

Venous thromboprophylaxis

Remission of AHA is often associated with high FVIII levels and, because patients are likely to have other risk factors for venous thrombosis, they should be assessed and treated with appropriate venous thromboprophylaxis (Huth-Kuhne et al, 2009).

Relapse

Relapse has been reported in 20% of 102 patients at a median of 7·5 months (range 1 week to 14 months) (Collins et al, 2007). An interim analysis of 176 evaluable patients in the EACH2 registry reported similar results with relapse in 19% of those treated first-line with steroids, 13% for steroids and cylcophosphamide and 0% in those treated with first-line rituximab (Collins et al, 2009). Patients require prolonged follow up and should be advised to report appropriate symptoms early. In our centre, the practice is to follow up indefinitely and should be for a minimum of 24 months.

Conclusions on inhibitor eradication

There is general consensus that immunosuppression aimed at eradicating the inhibitor should be started as soon as the diagnosis of AHA is made (Morrison & Ludlam, 1995; Hay, 1998; Delgado et al, 2003; Hay et al, 2006; Kessler & Asatiani, 2006; Huth-Kuhne et al, 2009). There are no convincing data to suggest that one immunosuppressive regimen is superior to any other or that the choice of regimen should be based on the inhibitor titre or FVIII level. First line therapy is at the discretion of the clinician based on the clinical circumstances and taking into account the potential side effects of each treatment option. Although the median time to CR is about 4–6 week, response demonstrated by a fall in inhibitor titre or increase in FVIII level, is usually seen earlier. If a patient does not respond to first line steroids then a cytotoxic agent or rituximab can be added. Similarly, if a patient fails first line rituximab then steroids and cytotoxics agents may be successful. Ciclosporin A is a useful second line option. A regimen based on high dose FVIII and immunoabsorption can be considered for patients with severe bleeding.

Future developments

Clinical progress in AHA is hampered by the small numbers of patients and difficulties in performing randomised studies. In the area of bleed control it is recognised that the haemostatic efficacy of all agents is unpredictable. A laboratory assay that has been clinically validated to predict successful haemostasis would be a very significant step forward. Understanding why the bleeding phenotype in AHA differs from congenital haemophilia may lead to a better understanding of the mechanism of haemostatic failure and possibly translate into improved haemostatic management. Access to new haemostatic agents is important and studies on the safety and efficacy of recombinant B-domain delete porcine factor VIII molecule and longer acting or enhanced rFVIIa molecules are awaited.

Studies in the field of inhibitor eradication are a major challenge, demonstrated by the fact that the literature contains only one randomised prospective clinical trial which was unable to recruit sufficient patients to provide interpretable data (Green et al, 1993). Trials that compare conventional steroid and cytotoxic agents with rituximab or investigate the role of FVIII would be useful. These trials will need to recruit hundreds of patients to be adequately powered and require international collaboration and significant resources to perform. It must be recognised that these trials may not be feasible and that registry data will need to be relied on for the foreseeable future.

Ancillary