Modern management of children with haemophagocytic lymphohistiocytosis.

Abstract

Haemophagocytic lymphohistiocytosis (HLH) is characterized by fever, hepatosplenomegaly, central nervous system symptoms, cytopenias, coagulopathy and lipid changes because of hypercytokinaemia and organ infiltration by phagocytosing histiocytes. Most cases are triggered by an infection. The common basis seems to be an inherited or acquired defect of immune effector cells. Diagnostic work-up should include the search for an infectious agent, especially viruses, and the measurement of disease markers such as fibrinogen, ferritin, triglycerides, sCD25, inflammatory cytokines and natural killer cell function. Initial therapy aims to suppress hyperinflammation, which can be achieved by immunosuppressive and cytostatic treatment with corticosteroids and cyclosporin A, and by etoposide. A combination of all three drugs is recommended for patients with severe symptoms, and cyclosporin A and corticosteroids are the treatment of choice for children with rheumatic diseases and HLH. Bone marrow transplantation, which offers a cure for nearly two-thirds of patients, should be performed in all children with primary genetic HLH. Haemophagocytic lymphohistiocytosis is not a single disease but can be encountered on the basis of a variety of underlying conditions leading to the same hyperinflammatory phenotype (Fig 1). Classification and underlying conditions of haemophagocytic lymphohistiocytosis. CHS, Chediak–Higashi syndrome; XLP, X-linked lymphoproliferative syndrome; FHLH, familial haemophagocytic lymphohistiocytosis; VAHS, virus-associated haemophagocytic syndrome; IAHS, infection-associated haemophagocytic syndrome; MAS, macrophage activation syndrome; chr.10, chromosome 10. Familial HLH, first described by Farquhar and Claireaux (1952) is an autosomal recessive disease that usually occurs in early infancy. Onset of the disease is in the first year of life in 70–80% of the cases (Janka, 1983; Arico et al, 1996). However, late-onset cases have also been reported (Allen et al, 2001). The disease proved to be invariably fatal when untreated (Janka, 1983). Cardinal symptoms are high fever, hepatosplenomegaly, cytopenias and various biochemical aberrations such as low fibrinogen and high triglycerides, ferritin, liver enzymes and bilirubin (Janka, 1983; Arico et al, 1996; Imashuku et al, 1997). Central nervous system (CNS) involvement is found in a substantial number of patients (Haddad et al, 1997). Organ infiltration by lymphocytes and benign-looking histiocytes with haemophagocytosis has given the disease its name. For a long time genetic markers for HLH were not known. Linkage analysis in four HLH families of Pakistani origin revealed a putative disease gene on chromosome 9q21.3–22 (Ohadi et al, 1999). Another region on chromosome 10q24 was then identified in linkage with the disease. The locus encodes for perforin, a protein that mediates the cytotoxic activity of natural killer (NK) and T cells. Perforin plays an important role in immune surveillance of virus-infected cells in the neonate and in early childhood and is involved in the downregulation of cellular immune activation (Stepp et al, 1999). Since then several investigators have reported different mutations but the mutation in amino acid (aa)374 appears to be most abundant (Clementi et al, 2001; Ericson et al, 2001; Feldmann et al, 2002; Suga et al, 2002). The overall frequency of perforin mutations is only between 20–30%. We found a high frequency of mutations in our Turkish patients (10 of 23) whereas in only three of 19 patients of German ancestry could mutations be identified. Successful prenatal diagnosis was reported in two families (Zur Stadt et al, 2002). Besides familial HLH, other well-described genetic immune deficiency syndromes, such as Chediak–Higashi syndrome (CHS) Griscelli syndrome and X-linked lymphoproliferative syndrome (XLP) can exhibit the clinical picture of HLH (Dufourcq-Lagelouse et al, 1999). Defective trafficking or degranulation of secretory granules in Griscelli syndrome and CHS leads to impaired cellular cytotoxicity (Dufourcq-Lagelouse et al, 1999). In XLP, in which the Epstein–Barr virus (EBV) can trigger fulminant HLH, dysfunction of the 2B4 receptor (CD244) because of mutations in the XLP gene SH2D1A results in the inability of NK cells to kill EBV-infected cells. A clinical picture resembling HLH has also been described for several inborn metabolic diseases (Mandel et al, 1990; Ikeda et al, 1998; Duval et al, 1999). Risdall et al (1979) described several patients, mostly adults after organ transplantation, the majority of whom had evidence of a viral infection and presented with similar clinical symptoms and findings as HLH. The authors designated this as virus-associated haemophagocytic syndrome (VAHS). Subsequently, it became clear that not only viruses could trigger the picture of HLH but occasionally also other organisms, such as bacteria, protozoa or fungi (Reiner & Spivak, 1988; Janka et al, 1998). Thus the term infection-associated haemophagocytic syndrome (IAHS) replaced VAHS. Not unlike familial HLH, IAHS proved to be a dangerous disease with a fatality rate of more than 50% in children (Janka et al, 1998). Whereas the presence of an infection was originally thought to discriminate between familial and acquired forms it is now clear that most episodes in the genetic forms of HLH are also triggered by infections (Henter et al, 1993; Arico et al, 1996). With a growing body of literature it also became apparent that most patients with IAHS did not have a known underlying genetic or iatrogenic immune deficiency but had been perfectly healthy before. In these cases the immune defect may be not apparent before the individual encounters a particular infectious organism. In 1985 the first report of a HLH-like picture in children with rheumatic disease was published (Hadchouel et al, 1985). The term macrophage activation syndrome (MAS) has been commonly used for episodes of HLH in rheumatic and other autoimmune diseases but it has recently been suggested that the term HLH should be used for this condition as well (Ramanan et al, 2002). Another secondary form of HLH may occur in association with malignancies, mostly of the haematopoietic system but also of various other types. HLH can develop before or during treatment, associated with an infection or without a known triggering factor. In cases with malignant lymphoma, such as angiocentric immunoproliferative lesion-like lymphoma or CD30-positive large cell anaplastic lymphoma, the malignancy is often masked and the patients are misdiagnosed as having VAHS (Janka et al, 1998). The term primary HLH is commonly used for children without underlying disease who either have a positive family history or presumably have a genetic defect due to their young age at presentation. All other forms are called secondary HLH. The clinical presentation of HLH is caused by a hyperinflammatory syndrome because of hypercytokinaemia of various pro-inflammatory mediators such as tumour necrosis factor α (TNF-α), interleukin (IL) 6, IL-8, IL-12, IL-18, macrophage inflammatory protein (MIP 1-α), interferon γ (INF-γ), and a number of haematopoietic growth factors released by stimulated lymphocytes and histiocytes (Henter et al, 1991a; Osugi et al, 1997; Teruya-Feldstein et al, 1999; Schneider et al, 2002). INF-γ attains much higher concentrations in plasma during active disease than in patients with sepsis or with autoimmune diseases. Remarkably, the inflammatory cytokine IL-1β is not elevated in classical HLH cases (Henter et al, 1996). Inflammatory cytokines are responsible for disease markers such as cytopenias, high triglycerides and coagulopathy. Markedly elevated levels of the alpha chain of the soluble IL-2 receptor (sCD25) and high concentrations of the soluble FAS (CD178) ligand as cell specific disease markers reflect stimulation of histiocytes and T cells (Komp et al, 1989; Imashuku et al, 1998; Schneider et al, 2002). Ferritin, which is also an important marker, is upregulated by oxidative stress (Tsuji et al, 2000) and may be related to a compensatory immune response in HLH. The uncontrolled activation of immune cells is probably because of an ineffective immune response triggered by exogenous stimulants such as infectious organisms and toxins or endogenous factors such as radical stress, tissue damage or metabolic products (Fig 1). The underlying immune defect which conditions the development of HLH can be an inherited defect of immune effector cells or can be acquired as an iatrogenic immune deficiency after transplantation or cytotoxic chemotherapy. An acquired immune deficiency may also result from infection of immune effector cells (Jerome et al, 1998) or indirectly through the immunosuppressive action of inflammatory mediators (Poggi et al, 1998). Patients with HLH exhibit impairment of their NK cells. NK cell function, measured as lysis of K562 cells in a standard 4 h chromium release assay, was impaired in all cases in our experience and that of others (Sullivan et al, 1988; Eife et al, 1989; Arico et al, 1996; Schneider et al, 2002) whereas not all patients reported in a Japanese study had decreased NK cell activity (Imashuku et al, 1998). The impairment is permanent in primary genetic forms (Eife et al, 1989; Schneider et al, 2002), providing a strong argument that this is not a secondary phenomenon but is the primary defect. However, patients with a persistent defect have also been identified in secondary, transient infection-associated forms (Janka et al, 2000). This suggests an inherited defect in these cases as well, or an infection-induced defect in cellular cytotoxicity. Impaired NK cell function may not only be due to an inherent defect of the killing machinery, such as perforin mutations, but may also be due to defective dendritic cell maturation. Sufficient evidence has been accumulated to indicate that dendritic cells control the differentiation and function of several NK cell populations (Yu et al, 2001). Modifications of the cytotoxicity assay revealed a residual cellular toxicity in some patients either in the presence of mitogen or IL-2 or after prolonged effector/target incubation time (Schneider et al, 2002). Type 3 NK cell deficiency, defined as total lack of cellular cytotoxicity unaffected by the above measures, is associated with the poorest prognosis (Schneider et al, 2002). A high percentage of parents and siblings of patients with HLH exhibit an impaired function of NK cells (Sullivan et al, 1998; Schneider et al, 2003). Decreased perforin expression in cytotoxic cells was recently reported for patients with systemic onset juvenile arthritis (Wulffraat et al, 2003) and for three of seven patients with the same disease complicated by macrophage activation syndrome (Grom et al, 2003). In the latter group, NK cell activity was also found to be impaired in all patients regardless of perforin expression. Defective NK and cytotoxic T-cell activity, either inherited or acquired, may be a common pathway in patients with HLH, leading to similar clinical manifestations. The classical symptoms of HLH are fever, hepatosplenomegaly, and cytopenias preceded by an upper respiratory or gastrointestinal infection in many patients. In some patients neurological symptoms such as seizures and cranial nerve palsies (Henter & Elinder, 1992) or hepatic failure (Parizhskaya et al, 1999) may be the first signs. Conspicuous laboratory findings, including elevated values of ferritin, triglycerides, liver enzymes, bilirubin, and lactate dehydrogenase (LDH) but low fibrinogen as well as haemophagocytosis, are initially only present in some of the patients but may be detected by repeated testing. Besides isolated hypofibrinogenaemia a more profound and sometimes life-threatening coagulopathy may develop (Hadchouel et al, 1985; Chen et al, 1998; Janka et al, 1998). All the above laboratory changes disappear upon successful treatment. There is no reliable way other than proven familial genetic disease or a known underlying disease to distinguish primary HLH from the secondary forms. Secondary cases are usually above 1 year of age at onset but otherwise presenting symptoms and laboratory findings overlap (Janka et al, 2001). A formal comparison of paediatric and adult HLH cases has not been attempted but from larger series in adults, including patients without and with known underlying conditions, such as rheumatic and malignant diseases, it appears that the same symptoms are found in both adults and children (Reiner & Spivak, 1988; Emmenegger et al, 2001). Triglycerides were usually not measured and there is more emphasis on elevated ferritin levels. In children with a presumed or proven genetic defect, the disease is usually rapidly fatal when untreated. Progressive cytopenias leading to bacterial or fungal sepsis, pneumonia and bleeding as well as cerebral dysfunction are the terminal events (Janka, 1983). Many patients initially show improvement with unspecific therapies, like transfusions or antibiotics, and some children may have prolonged intervals without the need for treatment between exacerbations of the disease (Janka-Schaub et al, 1999). Atypical familial cases, with late onset of the disease (Allen et al, 2001), even in adulthood, and perforin mutations (Clementi et al, 2002) have recently been reported. Considering the variety of underlying conditions which may lead to the clinical picture of HLH (Fig 1) a thorough history and diagnostic work-up is necessary in non-familial cases. If there is developmental delay, mental retardation or aminoaciduria, a metabolic disease should be considered. Albinism suggests CHS or Griscelli syndrome; giant granules in leucocytes are typical for the former. HLH in males associated with EBV infection or a positive family history for fatal infectious mononucleosis, agammaglobulinemia or lymphoma on the maternal side should prompt molecular analysis for XLP. If there is a history of rheumatic or other autoimmune disease, an associated macrophage-activation syndrome has to be considered. Lymphomas can present a formidable diagnostic problem as the lymphoma may be masked initially but, in our experience, it is a rare cause of HLH in children. The HLH is triggered by an infectious agent in most patients. This is also true for familial cases (Henter et al, 1993; Arico et al, 1996). Leading organisms are viruses, but bacteria, protozoa and fungi may also be involved (Reiner & Spivak, 1988; Janka et al, 1998). Among the herpes viruses, EBV is a frequent trigger for HLH especially in cases reported from Asia (Imashuku, 2002). Other viruses that have to be considered are cytomegalovirus, parvovirus B19, human herpes virus 6, adenovirus and varicella-zoster virus (Janka et al, 1998). Serological studies are less mandatory than the identification of antigen or viral DNA by polymerase chain reaction. Whereas bacterial and fungal infections are usually considered in a febrile patient, leishmaniasis, which is a fairly frequent cause of HLH, can be easily overlooked. In 1991 the familial haemophagocytic lymphohistiocytosis (FHL) Study Group of the Histiocyte Society proposed diagnostic guidelines for HLH (Henter et al, 1991b). They include: fever of ≥7 days duration, splenomegaly, cytopenia affecting at least two cell lines (haemoglobin < 9·0 g/dl, platelets < 100 × 109/l, neutrophils < 1·0 × 109/l), hypofibrinogenaemia (<1·5 g/l) and/or hypertriglyceridaemia (>2 mmol/l) as well as haemophagocytosis in bone marrow or other tissues (Henter et al, 1991b). Triglycerides can rise during the course of infections but, apart from patients with bacterial sepsis, usually do not surpass values of 3 mmol/l in children with febrile illnesses (Garbagnati, 1993). Of 65 patients treated in Germany, 69% had triglyceride levels >3 mmol/l and 35% had levels >5 mmol/l. Low fibrinogen is a rather specific marker for HLH, as patients with infections (apart from sepsis with disseminated intravascular coagulation) are expected to have normal or elevated fibrinogen values. Sensitivity, however, is low; only 53% of our patients had a fibrinogen level below 1·5 g/l. Ferritin may be elevated in children with infections but, apart from patients with human immunodeficiency virus, values remain below 200 μg/l (Cemeroglu & Özsoylu, 1994; Dimitriou et al, 2000). Taking into account that newborns already have higher values without infections (Dimitriou et al, 2000) a ferritin value above 500 μg/l in a febrile child (sensitivity 69% in the cohort of German patients) is a strong indication of HLH. However, it should be remembered that children with systemic juvenile rheumatoid arthritis without signs of MAS might have ferritin values up to 10 000 μg/l (Pelkonen et al, 1986). The marker that showed the highest sensitivity (93%) was a sCD25 value above 2400 U/ml; its specificity compared with sepsis or other infections was 1·0 (http://www.histio.org/society/HLH/janka1.shtml). Even higher levels have been reported in leukaemias (Komp et al, 1989) and lymphomas (Perez-Encinas et al, 1998). Although NK cell activity quantification is usually not available in time for treatment decision and is reserved for specialized laboratories, it should be included as an important diagnostic marker. In parents of suspected genetic cases restricted NK cell function is a frequent finding and helps to confirm familial disease (Sullivan et al, 1998; Schneider et al, 2003). Bone marrow aspiration should be performed in all cases. The marrow usually has normal cellularity and increased erythropoiesis, which, however, is ineffective, as reflected by an inadequate reticulocyte count in the presence of anaemia. If positive for haemophagocytosis, increased numbers of histiocytes are seen engulfing mainly red cells and erythroblasts, but also platelets and white cells. In the patients from Germany, haemophagocytosis was present in only 32% of cases in the first bone marrow and in 85% at diagnosis. A spinal tap is mandatory as CNS involvement, evidenced by moderate pleocytosis and/or elevated spinal fluid protein levels, is present in a substantial number of patients, most of whom have no neurological symptoms (Haddad et al, 1997). Magnetic resonance imaging is strongly recommended in patients with signs of CNS involvement. Table I presents the revised diagnostic criteria of the HLH study group of the Histiocyte Society, which has now been adopted by the Society. Perforin expression in NK cells and cytotoxic lymphocytes measured by flow cytometry has proven to be of value in identifying patients with mutations in the perforin gene (Arico et al, 2002; Kogawa et al, 2002) and should be performed in all patients. It may also be of interest in patients with systemic juvenile arthritis for whom decreased perforin expression and suppressed NK cell function was recently reported (Grom et al, 2003; Wulffraat et al, 2003). Mutation analysis of the perforin gene should be performed in all patients with negative perforin expression. Male patients with EBV-associated HLH should be considered for SH2D1A mutation analysis (Arico et al, 2001). Genetic counselling is important in all patients with HLH and prenatal diagnosis should be offered to families with known genetic mutations. Familial (primary) HLH used to be a rapidly fatal illness with most children succumbing to the disease within 6 months. Various attempts to control the disease including corticosteroids, exchange transfusions, cytostatic drugs, and splenectomy resulted in only transient improvements in most of the patients. Occasional long-term survivors most probably had the non-genetic form of HLH. In 1980 the epidophyllotoxin etoposide was introduced for treatment of HLH (Ambruso et al, 1980) and in some small series excellent responses with etoposide either as single agent or in combination with steroids were reported (Blanche et al, 1989; Janka, 1989). However, relapses were frequent (Janka, 1989). Cranial irradiation was used together with etoposide, steroids and intrathecal methotrexate by Fischer et al (1985), the same group later reported remissions with antithymocyte globulin, steroids and cyclosporin A (Stephan et al, 1993). Cyclosporin A also proved to be effective as a single agent for maintenance (Loechelt et al, 1994) and in children with rheumatic diseases and MAS (Mouy et al, 1996). Immunoglobulins were shown to be beneficial in some older children with infection-associated HLH (Freeman et al, 1993). The immediate aim of treatment is to suppress the severe hyperinflammation that is responsible for the life-threatening symptoms. Fever, pancytopenia, liver damage and CNS symptoms can all be attributed to the action of cytokines released by activated lymphocytes and histiocytes. The second aim is to kill pathogen-infected antigen-presenting cells and thus remove the stimulus for the ongoing but ineffective activation of NK and T-suppressor cells. In genetic cases the ultimate aim must be to exchange the defective immune system by normally functioning immune effector cells. Corticosteroids are not only cytotoxic for lymphocytes but also inhibit production of CD95 ligand and differentiation pathways of dendritic cells (Woltman et al, 2002) as well as the expression of chemokines and cytokines (Galon et al, 2002). Cyclosporin A exerts its immunosuppressive action by interfering with the cyclophilin pathway (Fruman et al, 1994). Immunoglobulins act by cytokine- (Ross et al, 1997) and pathogen-specific antibodies, which in turn stimulate Fc receptor function. Absorption of cytokines to erythrocytes may explain why several patients with HLH improve transiently after red cell or platelet transfusions. Etoposide is a cytotoxic drug with high activity in monocytic and histiocytic diseases. Etoposide was shown to inhibit the synthesis of EBV nuclear antigen (EBNA) (Kikuta & Sakiyama, 1995) and may thus prevent clonal expansion of EBV-infected T cells in patients with HLH. It also seems to eliminate virus-infected immune cells. The ultimate aim of restoring immune function permanently in genetic cases can, however, only be achieved by stem cell transplantation (SCT). Children with HLH have such a broad spectrum of underlying conditions that uniform treatment cannot do justice to all patients. Underlying disease(s), the age of the patient, presence of familial disease, identification of an infectious organism and especially the severity of symptoms have all to be considered. Table II gives treatment recommendations based on the HLH protocol (Henter et al, 1997) and on our own experience. The rare patient with lysinuric protein intolerance should be distinguishable by a more chronic and intermittent course and will usually not need immunosuppressive treatment. Corticosteroids and cyclosporin A were beneficial in one case with several severe episodes of HLH (Bader-Meunier et al, 2000). Children with the accelerated phase of CHS or Griscelli syndrome as well as EBV-associated HLH in patients with XLP respond to etoposide-containing regimens (Bejaoui et al, 1989; Gurgey et al, 1996; Kumar et al, 2001) but as genetic defects they are only curable by SCT. In children with systemic inflammatory rheumatic disorders in whom MAS/HLH may be a life-threatening and sometimes even fatal complication, immunosuppressive treatment is mandatory. As in familial HLH at presentation the diagnostic criteria proposed by the Histiocyte Society (Henter et al, 1991b) are not fulfilled in a substantial number of patients (Hadchouel et al, 1985; Mouy et al, 1996; Stephan et al, 2001), and physicians are cautioned not to allow delay in treatment because of lack of haemophagocytosis (Stephan et al, 2001). The recommendation for HLH associated with rheumatic diseases in children is to discontinue non-steroidal antiphlogistic medication and to treat with cyclosporin A either as first line drug or as second line drug if no rapid improvement is achieved with high dose (≥2 mg/kg bw) corticosteroids (Mouy et al, 1996; Stephan et al, 2001). Whereas immunoglobulins failed in children with MAS (Stephan et al, 2001) responses were observed in adult patients, either as a single agent (Emmenegger et al, 2001) or with concomitant drugs, mostly corticosteroids (Larroche et al, 2000). No firm treatment recommendations can be given for children with HLH during the course of a malignant disease. In severe cases treatment according to primary HLH with etoposide and steroids may be appropriate (Janka et al, 1998). The most difficult question is how children with non-familial HLH should be treated especially if an infectious agent has been identified. There are no reliable criteria to differentiate genetic versus non-genetic forms, including symptoms, infections, and laboratory values (Janka et al, 2001). Age above 1 year is more compatible with a secondary form but familial cases with late onset have been reported. If there is proof of an infection and treatment is available, as is the case for cytomegalovirus and some other pathogens, the appropriate drugs should be given. However, in most cases treatment of the infection is not sufficient to abrogate hypercytokinaemia and its deleterious effects, with the possible exception of leishmaniasis in which therapy with liposomal amphotericin B is usually sufficient. The decision whether to treat a child with presumed secondary HLH with immunosuppressive, immunomodulatory, or even cytostatic drugs should depend solely on the clinical condition of the patient and associated laboratory changes. Patients who are critically ill with unremitting fever, progressive bi- or pancytopenia, clotting abnormalities, or marked signs of liver damage, and also patients below 1 year of age should be started on the HLH protocol. The risk of etoposide, even if given for 8 weeks, is by far exceeded by the risk of loosing a child through inadequate treatment. According to the authors’ experience in children with mild to moderate symptoms, treatment may be withheld or immunosuppressive and immunomodulatory drugs such as corticosteroids, immunoglobulins or cyclosporin A may be given initially. There are very few published reports about children with HLH responding to such treatment and with one exception all were older than 1 year of age (Goulder et al, 1990; Freeman et al, 1993; Chen et al, 1998; Erduran et al, 2000). It has to be kept in mind that many children with secondary forms may start out with mild symptoms that evolve into a life-threatening course, especially in EBV associated cases (see below). The EBV-related HLH, either because of primary infection or reactivation, is a frequent and serious problem in Asia (Su et al, 1990; Imashuku, 1997, 2002; Chen et al, 1998). Fatal cases have also been reported in Western countries (Mroczek et al, 1987). Eighteen of the 126 HLH cases in the German registry were caused by EBV and six were fatal (Beutel et al, 2002, updated 2003). A higher percentage of paediatric cases (15 of 82) reported from Japan were EBV-associated and six patients died (Imashuku et al, 1997). In most cases monoclonal proliferation of EBV-infected T or NK cells has been found (Kawaguchi et al, 1993). The progression to lymphoma or leukaemia seems to be a special problem in these cases (Chen et al, 1998; Imashuku et al, 1999a). Genetic or environmental factors may play a role in the high prevalence of EBV-HLH cases in Asia. Recently it was shown that the prognosis for EBV-related HLH has improved considerably with immunochemotherapy including etoposide and corticosteroids (Chen et al, 1998; Imashuku et al, 1999b), and therapy according to the HLH-94 protocol was suggested as the initial treatment of choice (Imashuku et al, 1999b). The same group later reported 47 children and young adults with EBV-associated HLH, including the 16 patients from the previous report. Twenty-one patients were first treated with corticosteroids alone, intravenous immunoglobulins alone, cyclosporin A alone, or a combination of these drugs (conventional regimens) without etoposide, and 26 received an etoposide-based regimen initially. Only one of the conventional regimens induced a complete remission leading to a switch to etoposide in 17 patients. In a multivariate analysis, early introduction of etoposide-based regimens was the only significant variable for survival (relative risk of death 14·1 times higher for patients not receiving etoposide or receiving it later than 4 weeks after diagnosis). Higher doses of etoposide during the first 4 weeks of treatment were also associated with a better prognosis. The additional use of cyclosporin A early on appeared to be linked to a higher survival rate but this was not statistically significant (Imashuku et al, 2001). Although it remains to be determined whether etoposide is needed for all patients with EBV-associated HLH (six of 18 patients in the German registry recovered without etoposide) or could be restricted to moderately severe and severe cases, it is our opinion that the indication for its use should be handled rather generously as initially mild cases may rapidly progress and the risk to be faced with a life-threatening situation outweighs the risks of etoposide. The FHL study group of the Histiocyte Society was formed in 1989 and a treatment protocol was devised on the basis of existing experience (Ambruso et al, 1980; Janka, 1989; Stephan et al, 1993; Loechelt et al, 1994). Etoposide and corticosteroids were chosen as active agents for first-line treatment. Because of better penetration into the cerebrospinal fluid (CSF), dexamethasone was preferred to prednisolone. The initial intensive 8-week treatment period was followed by maintenance with cyclosporin A reinforced by alternating pulses of etoposide and dexamethasone. Patients with a family history or presumed familial disease because of either incomplete response or relapses were kept on maintenance until SCT, whereas in children without family history and complete resolution of symptoms, therapy was to be selectively stopped after 8 weeks. Relapses after termination of treatment qualified the patient for renewal of therapy and SCT (Henter et al, 1997). Between 1 July 1994 and 30 June 1998, 119 children below 15 years of age who fulfilled the diagnostic criteria or were familial cases were entered into protocol HLH-94. The patients were recruited from 21 countries. After a median follow-up of 37·5 months 63 of 113 children were alive with a probability of 55% for survival at 3 years for all cases and 51% for proven familial cases. Forty patients underwent SCT, no familial case survived without a transplant. There were 23 patients alive (20 off therapy) who had not undergone SCT. Their mean age at diagnosis was 47, compared with 13 months for familial cases, and only one patient was below 1 year of age. Presumably these patients in retrospect had non-genetic secondary HLH but were sufficiently ill to require intensive therapy. Not considering these 23 probably secondary cases, 65 of 90 (72%) patients had a SCT after chemoimmunotherapy of whom 40 were still alive at the time of writing (Henter et al, 2002). The HLH-94 protocol was a big step forward in showing that systemic treatment can effectively improve and even control clinical symptoms of HLH in the majority of patients who can then receive curative treatment with SCT. However, a considerable number of patients fail treatment either as non-responders or with recurrent disease on therapy. As the number of patients is small, even with an international protocol, it is difficult to devise strategies to optimize treatment. In protocol HLH-94 there was a weak link after 4 weeks, where dexamethasone was already reduced and cyclosporin A not yet started. Several patients relapsed in this period. Consequently in the next international HLH protocol, which is in preparation, cyclosporin A will be moved up-front. Also worrisome are the patients who do not respond in the first 4 weeks. For these patients, alternative therapies have to be found and the molecular basis for inefficient therapy has to be clarified. More reliable prognostic and response criteria are also urgently needed. At present it is unclear how and when best to use granulocyte colony stimulating factor (G-CSF). Responding patients quickly recover their platelet counts but neutropenia may persist for longer periods. In these patients, G-CSF may be beneficial whereas it is probably ineffective in florid disease. Substantial numbers of patients have an elevated CSF cell count or protein, but only a minority of those exhibit neurological symptoms (Haddad et al, 1997). The HLH protocol recommends systemic treatment alone initially, to repeat the lumbar puncture after 2 weeks of dexamethasone, and to use intrathecal therapy only in patients without neurological improvement or with persistent signs of CSF inflammation after 2 weeks. In the 35 patients with CNS symptoms at diagnosis treated according to the HLH study, an equal percentage had normalization of symptoms with or without intrathecal methotrexate (Henter et al, 2002). Although the value of intrathecal methotrexate has thus not been proven it can be assumed that the more severely affected patients received this treatment. Also from our own experience intrathecal methotrexate was beneficial in patients with neurological symptoms. In 1986 the first report of successful bone marrow transplantation in HLH from an HLA-identical sibling was published (Fischer et al, 1986). Subsequently, several single cases and larger series with transplants from either related or unrelated donors confirmed that permanent control of the disease was possible (Blanche et al, 1991; Baker et al, 1997; Jabado et al, 1997; Dürken et al, 1999; Ardeshna et al, 2001). The survival rate of 62% after BMT, as reported in a recent review (Dürken et al, 2001), is identical to the 62% 3-year survival probability for children treated with the HLH-94 protocol (Henter et al, 2002). Only a minority of the patients had been transplanted from HLA-identical siblings. The preparative regimen, usually consisting of busulphan, cyclophsophamide and etoposide with additional antithymocyte globulin for unrelated donor transplants, is also recommended in protocol HLH-94 (Henter et al, 2002). Survival was worse for patients transplanted with active disease in one centre (Baker et al, 1997) but not in another (Dürken et al, 1999). NK cell function was reported to return to normal between 3 and 12 months after BMT (Baker et al, 1997; Jabado et al, 1997). Mixed chimerism after BMT is encountered in some patients but complete abrogation of host haematopoiesis does not seem to be necessary for long-term survival (Landman-Parker et al, 1993; Jabado et al, 1997; our unpublished observations). The availability of matched sibling donors for HLH is limited and there is also concern that the donor may harbour the disease. Although HLH in siblings usually manifests itself around the same time of life, the age of onset was markedly different in several familial cases (Arico et al, 1996; Allen et al, 2001; Ardeshna et al, 2001). Two cases have been reported in which transplants were received from a sibling who was later affected, where there was a relapse of HLH 1 and 3 years after BMT (Blanche et al, 1991; Ardeshna et al, 2001). Absence of a matched donor should not exclude patients from transplantation. Of 14 patients who had a T-cell depleted marrow transplant from HLA non-identical related donors, 11 achieved a sustained engraftment (Jabado et al, 1997). In 10 there was a full haplotype mismatch between donor and recipient. An anti-LFA-1 and an anti-CD2 antibody were added to the conditioning regimen. Nine patients are alive and well. The experience with haploidentical family donors was less favourable in study HLH-94 with only six of 14 survivors (Henter et al, 2002). Stem cell transplantation in six refractory/aggressive EBV-related HLH cases was reported from Japan. All children had non-familial disease and were above 2 years of age. Five cases were confirmed to have clonally proliferating EBV-infected T or NK cells (Imashuku et al, 1999a). Relapses have been reported with rejection of the graft (Arico et al, 1996; Baker et al, 1997; Jabado et al, 1997; Ardeshna et al, 2001) but this is exceedingly rare after transplants with matched siblings or unrelated donors. Symptoms of HLH after transplantation do not necessarily herald the relapse of the original disease; after BMT patients are in an immunocompromised state and may develop new infection-triggered HLH (Reardon et al, 1991). Myeloablative therapy and autologous SCT was suggested for patients without suitable donor by Ohga et al (1997) who reported a patient with a remission of 2·5 years duration after this procedure. However, the patient received cyclosporin A and corticosteroids after BMT, which can produce long remissions. Moreover, this approach is not curative in familial cases. The HLH is a disease with many facets regarding clinical presentation as well as underlying diseases. It should be considered in every patient with unremitting fever, hepatosplenomegaly and cytopenias. Although awareness for this condition has increased in recent years, HLH is still overlooked in many cases, especially in older children. The proof of an infectious agent does not help to discriminate between primary, genetic forms and secondary, non-genetic forms. As effective treatment is available, adequate diagnostic management leading to early diagnosis is mandatory. Better prognostic parameters and alternative drugs for patients without response to standard treatment are needed. We would like to thank the Foerdergemeinschaft Kinderkrebszentrum Hamburg e.V. and the Histiozytosehilfe e.V. for their financial support of research on the pathophysiology of HLH.