ARTICLES

High-dose cyclophosphamide in severe aplastic anaemia: a randomised trial

John F Tisdale, Daniel E Dunn, Nancy Geller, Michelle Plante, Olga Nunez, Cynthia E Dunbar, A John Barrett, Thomas J Walsh, Stephen J Rosenfeld, Neal S Young

Summary
Background High-dose cyclophosphamide has been proposed as an alternative immunosuppressive agent for treatment of severe aplastic anaemia, with a response rate similar to that with regimens containing antithymocyte globulin (ATG) but neither relapse nor clonal haematological complications. We undertook a phase III, prospective, randomised trial to compare response rates to immunosuppression with either high-dose cyclophosphamide plus ciclosporin or conventional immunosuppression with ATG plus ciclosporin in previously untreated patients. Methods Between June, 1997, and March, 2000, 31 patients were enrolled. 15 were assigned cyclophosphamide (1 h intravenous infusion of 50 mg/kg daily for 4 days) and 16 were assigned ATG (40 mg/kg daily for 4 days); both groups received ciclosporin, initially at 12 mg/kg daily with adjustment to maintain concentrations at 200400 g/L, for 6 months. The primary endpoint was haematological response (no longer meeting criteria for severe aplastic anaemia). The trial was terminated prematurely after three early deaths in the cyclophosphamide group. Analyses were by intention to treat. Findings Median follow-up was 219 months (range 133). There was excess morbidity in the cyclophosphamide group (invasive fungal infections, four cyclophosphamide vs no ATG patients; p=0043) as well as excess early mortality (three deaths within the first 3 months cyclophosphamide vs no ATG patients; p=0101). There was no significant difference at 6 months after treatment in the overall response rates among evaluable patients (six of 13 [46%] cyclophosphamide vs nine of 12 [75%] ATG). Interpretation A longer period of observation will be necessary to assess the secondary endpoints of relapse and late clonal complications as well as disease-free and overall survival. However, cyclophosphamide seems a dangerous choice for treatment of this disorder, given the good results achievable with standard therapy. Lancet 2000; 356: 155459 See Commentary page ????

Introduction
Bone-marrow failure due to aplastic anaemia can be effectively treated with immunosuppressive therapy.1,2 For most patients with severe disease, as defined by peripheral blood counts, the combination of an antilymphocyte serum such as antithymocyte globulin (ATG) and ciclosporin results in clinically worthwhile haematological responses, producing improvement in the neutrophil count sufficient to prevent infection and obviating the need for transfusion of red blood cells and platelets.2,3 Prognosis is much improved by response to immunosuppressive therapy; overall survival rates are similar to those achieved with allogeneic stem-cell transplantation, especially in older patients and those with moderate neutropenia.4 Nevertheless, there are problems associated with immunosuppression.1 First, incomplete recovery of blood counts is common, although in many cases not clinically important. Second, many patients relapse, requiring repeated or extended treatment with ciclosporin or further courses of ATG.5 Third, other haematological disorders evolve from aplastic anaemia after apparent recovery from initial marrow failure.6 Paroxysmal nocturnal haemoglobinuria was observed in 20% of patients in a US study,7 and in the latest analysis of the large European experience, the probability of paroxysmal nocturnal haemoglobinuria, myelodysplasia, or acute leukaemia was 17% at 10 years.6 Finally, the cost of a course of antilymphocyte globulin is prohibitive in less developed countries, where aplastic anaemia is more common than in Europe or the USA. The usefulness of ATG in marrow failure was discovered in the setting of bone-marrow transplantation; some patients recovered their own haemopoietic function when engraftment failed.8 Similar autologous reconstitution occurred after conditioning with cyclophosphamide for transplantation and, in one case report, in the absence of marrow grafting.9,10 At Johns Hopkins University Hospital, a small series of patients with aplastic anaemia was deliberately given high-dose cyclophosphamide over a 10-year period during which ATG was not always available from its sole US supplier.11 Of ten patients who received the drug at a dose of 45 mg/kg daily for 4 consecutive days (three also received concomitant ciclosporin), seven responded; among the six long-term survivors, all ultimately had normal blood counts, with no relapses or evidence of a second haematological complication. The investigators suggested that efficacy was secondary both to cyclophosphamides more intensive immunosuppressive action and to eradication of premalignant clones through its direct cytotoxicity. On the basis of these results, high-dose cyclophosphamide has been promoted as curative therapy in aplastic anaemia, capable of inducing durable and complete haematological responses without risk of late complications,12 although the small pilot study was not sufficiently powered to address these potential late complications. To assess the role of cyclophosphamide in

Hematology Branch, National Heart, Lung, and Blood Institute, National Institutes of Health, Bethesda, MD, USA (J F Tisdale MD, D E Dunn MD, N Geller PhD, M Plante PharmD, O Nunez BSN, C E Dunbar MD, A J Barrett MD, T J Walsh MD, S J Rosenfeld MD, N S Young MD) Correspondence to: Dr John Tisdale, Molecular and Clinical Hematology Branch, National Institute of Diabetes and Digestive and Kidney Disorders, Building 10, Room 9N116, National Institutes of Health, Bethesda, MD 20892, USA (e-mail: johntis@intra.niddk.nih.gov)

1554

THE LANCET Vol 356 November 4, 2000

ARTICLES

the treatment of severe aplastic anaemia, we initiated a prospective, randomised comparison of standard treatment (ATG and ciclosporin) and treatment with high-dose cyclophosphamide and ciclosporin. The primary endpoint was the haematological response rate.

31 randomised

Methods
Study participants The protocol was approved by the Institutional Scientific Review Committee and the Institutional Review Board of the National Heart, Lung, and Blood Institute. All patients gave written informed consent. We treated all referred patients over age 18 years (later changed to age 16) with severe aplastic anaemia, defined by bone-marrow cellularity of less than 30% (excluding lymphocytes) and at least two of the following features: absolute neutrophil count less than 050109/L, platelet count below 20109/L, or absolute reticulocyte count less than 60109/L. Previous immunosuppressive therapy was an exclusion criterion. Patients who had suitable histocompatible donors were referred for consideration of allogeneic bone-marrow transplantation; if they were unwilling to undergo transplantation they were considered for randomisation. We excluded patients with depressed cardiac function (cardiac ejection fraction <45% by resisting ventriculogram), Fanconis anaemia or other congenital bone-marrow-failure syndromes, or evidence of a clonal haematological disorder by cytogenetic analysis. Design and procedures Patients screened for eligibility and meeting study entry criteria were randomly assigned initial treatment with either cyclophosphamide plus ciclosporin or ATG plus ciclosporin. The Clinical Center Pharmaceutical Development Section (PDS) assigned treatment through a table of random numbers with variable block sizes. Differences in side-effects and their management between treatment groups precluded masking of the study. Patients received cyclophosphamide as an intravenous infusion over 1 h at 50 mg/kg daily on 4 consecutive days with concomitant mesna and vigorous intravenous hydration as prophylaxis against haemorrhagic cystitis. ATG was given at a dose of 40 mg/kg daily for 4 consecutive days. Oral prednisone (1 mg/kg daily) was started at least 6 h before the first dose of ATG as prophylaxis against serum sickness and was continued for 10 days, followed by tapering over the next 7 days. Ciclosporin was administered to all patients from day 1, initially at 12 mg/kg daily by mouth in divided doses every 12 h (the protocol was slightly modified before the last two patients were enrolled such that the start of ciclosporin treatment was delayed by 2 weeks). Drug dosing was adjusted to maintain a whole-blood ciclosporin concentration of 200400 g/L, and the drug was administered for at least 6 months. The dose was tapered by 5% per week over 20 weeks at the end of the 6month treatment period in responding patients. Nonresponders discontinued ciclosporin 6 months after ATG or cyclophosphamide was started. Antimicrobial prophylaxis with norfloxacin, fluconazole, and aciclovir (for patients seropositive for herpes simplex virus) started on day 1 and continued until the neutrophil count exceeded 050109/L. Aerosolised pentamidine was used as prophylaxis against Pneumocystis carinii monthly for the duration of ciclosporin treatment. Transfusions of red blood cells were given to maintain haemoglobin concentrations above 7 g/dL. Platelet transfusions were given to maintain platelet counts above

15 assigned cyclophosphamide + ciclosporin

16 assigned ATG + ciclosporin

15 evaluable at 3 months

14 evaluable at 3 months

13 evaluable at 6 months

12 evaluable at 6 months

Figure 1: Trial profile

10109/L. Patients refractory to platelet transfusions were not given them routinely, but serious bleeding was treated with transfusions irrespective of the platelet count. All blood products were irradiated to inactivate donor lymophocytes potentially capable of mediating graftversus-host disease, and were depleted of leucocytes by filtration to limit allosensitisation. Parenteral broadspectrum antibiotics were given if fever occurred during the period of neutropenia. If fever recurred or persisted on broad-spectrum antibiotics during neutropenia, amphotericin B was given. Granulocyte-colonystimulating factor (G-CSF) was not routinely administered but was allowed at the clinicians discretion for treatment of fever and persistent neutropenia. Granulocyte transfusions from G-CSF-mobilised normal donors were administered to patients with documented or suspected fungal disease and persistent fever despite treatment with amphotericin B. The primary endpoint of the study was haematological response rate. Analyses were by intention to treat. Secondary endpoints were overall survival, event-free survival, response duration, and evolution to paroxysmal nocturnal haemoglobinuria, myelodysplasia, or acute leukaemia. The definition of response was that the patients no longer met diagnostic criteria for severe aplastic anaemia; this definition is closely associated with independence from transfusion and improved survival.3 To define differences between treatment groups more closely, outcome was further classified by ordered, mutually exclusive response categories: partial responses with continued dependence on transfusions; partial response with transfusion independence and subnormal blood counts; and complete response with normal or near-normal blood counts (ie, absolute neutrophil count 100109/L, haemoglobin 10 g/dL, and platelet count 100109). Non-responders were offered cross-over to the other treatment at 6 months after the start of therapy but were classified as non-responders for the primary endpoint. Response was assessed at scheduled visits at 3, 6, and 12 months and then yearly. Improvement in blood counts dependent on exogenously administered growth factors was not regarded an meeting response criteria. Statistics A sample size of 91 patients per group was planned to allow the detection of a 20% difference in the response rate, from 60% to 80%, with 80% power in a two-sided

THE LANCET Vol 356 November 4, 2000

1555

ARTICLES

test at significance 005. This sample size would also have given a power of just under 90% to detect a difference in response of 70% in one group and 90% in the other. An interim analysis was planned to assess whether the treatments differed after half of the patients had been evaluated for response with a nominal significance level of 001. If no difference was seen, the trial would continue to completion and the final analysis would be undertaken at significance of 00455. This plan would maintain an overall significance of 005. Summary statistics were calculated overall and by treatment group. The Mann-Whitney-Wilcoxon test was used to compare the treatment groups in terms of age at diagnosis, baseline blood counts, and ordered response categories as assessed at 3 months and 6 months for all evaluable patients. Categorical variables (use of G-CSF, fungal infection, death within 3 months, and fungal infection or death within 3 months) were compared by use of Fishers exact test. The p value was calculated by summing the probabilities of all tables at least as extreme as that given by our data. Distributions of days on intravenous antibiotics, days on amphotericin B, days in hospital, days to reach an absolute neutrophil count of more than 050109/L, days of absolute neutrophil count below 020109/L or below 005109/L, number of transfused units of red blood cells, number of platelet transfusion events over the first 6 months, and overall survival distributions were estimated by the Kaplan-Meier method, and the treatments were compared by use of the log-rank test. Patients who died before 6 months and those not followed up to 6 months were censored in the analysis of transfused red blood cells and platelets. Statistical tests were two-sided. The statistical package STATA (version 5.0) was used for all analyses.

Measure Days in hospital Days of intravenous antibiotics Days on amphotericin B RBC units Platelet transfusions G-CSF Days with ANC <005109/L Days with ANC <020109/L Days to ANC >050109/L

Cyclophosphamide (n=15) 59 47 25 36 32 14 33 40 53

ATG (n=16) 12 7 0 11 9 3 0 0 3

p* 00009 00015 0030 0024 0033 0001 00014 00007 00029

RBC=red blood cells; ANC=absolute neutrophil count. *Log-rank test except for number receiving G-CSF (Fishers exact test). Number of patients treated.

Table 2: Median values for supportive care measures

Results
Accrual to the trial began in June, 1997 (figure 1). After 31 patients had been accrued, a third unexpected death occurred in the cyclophosphamide group within 6 weeks of the start of treatment. Although there was no stopping rule for excess mortality, these events prompted a report in March, 2000, to the Institutional Review Board. Both morbidity and mortality were reviewed, and the Board concurred with the investigators decision to stop accrual to the trial and to report the early results. All 31 patients had received the full course of assigned treatment. The two randomised groups were similar in terms of age at diagnosis and baseline absolute reticulocyte and platelet counts (table 1). However, patients in the cyclophosphamide plus ciclosporin group had a higher initial absolute neutrophil count than patients in the ATG plus ciclosporin group. Three patients assigned cyclophosphamide and five assigned ATG had a detectable population of cells deficient in glycosylphosphoinositol-anchored protein at presentation. The median period of observation at the time of analysis was 667 days (219 months). Compared with the
All (n=31) Cyclophosphamide (n=15) ATG (n=16) 475 (1867) 040 (001298) 224 (296577) 135 (725)

ATG group, patients assigned cyclophosphamide spent significantly longer in hospital, experienced longer duration of neutropenia, and required significantly longerduration treatment with broad-spectrum parenteral antibiotics and amphotericin B. Furthermore, those assigned cyclophosphamide needed more transfusional support with red blood cells and platelets during the first 6 months of follow-up. The median duration of absolute neutrophil count below either 020109/L or 005109/L was 40 and 33 days, respectively, in the cyclophosphamide group versus 0 days in the ATG group (table 2). The Kaplan-Meier estimates of the proportion of patients achieving neutrophil count of 050109/L or higher for both treatment groups are shown over time from treatment in figure 2. Although all but one patient in the cyclophosphamide group received G-CSF (compared with only three of 16 in the ATG group), a significant delay in neutrophil recovery was seen in the cyclophosphamide group; in fact all but one patient in the ATG group reached the target neutrophil count before the first patient in the cyclophosphamide group (log-rank test, p=00029). Before neutrophil recovery, four patients in the cyclophosphamide group developed documented invasive fungal infections (pulmonary aspergillosis in two, sinoorbitol aspergillosis in one, and sinus alternariosis in one). One patient died on day 42 (age 52, presenting absolute neutrophil count 015109/L); two needed aggressive surgical debridement; and three needed granulocyte transfusion. No similar events occurred among patients
10 Probability of reaching ANC >050 109/L

08

06

04

02

Cyclophosphamide ATG/ciclosporin

0 0
Number at risk Cyclophosphamide 15 ATG 16

Age at diagnosis (years) 42 (1867) 35 (1867) 047 (001298) 086 (015290) ANC (109/L)* 9 220 (296577) 220 (330572) ARC (10 /L) 11 (247) Platelet count (109/L) 13 (247)

20

40

60

80

100 120 140

Time (days)
15 5 12 2 4 1 4 1 4 1 0 1 0 1

ANC=absolute neutrophil count; ARC=absolute reticulocyte count. *One patient assigned cyclophosphamide and two patients assigned ATG presented with supersevere disease (ANC <020109/L).

Table 1: Median (range) values for baseline characteristics of patients

Figure 2: Kaplan-Meier estimates of proportion of patients reaching an absolute neutrophil count (ANC) of 050109/L

1556

THE LANCET Vol 356 November 4, 2000

ARTICLES

100 90 80 70 Response (%) 60 50 40 30 20 10 0

CR PRi PRd

Cy

ATG 3 months

Cy

ATG 6 months

Figure 3: Response proportions for evaluable patients at 3 and 6 months

Cy=cyclophosphamide; CR=complete response; PRi=partial response, transfusion-independent; PRd=partial response, transfusion-dependent.

assigned treatment with ATG (p=0043), Two other patients in the cyclophosphamide group were suspected to have fungal lung infections and were treated accordingly; one died on day 39 (age 35, presenting absolute neutrophil count 044109/L). Refractory sepsis developed in another patient in the cyclophosphamide group resulting in death on day 39 (age 65, presenting absolute neutrophil count 15109/L). In all, three patients treated with cyclophosphamide died within 3 months, compared with none treated with ATG (p=0101). Altogether, six cyclophosphamide-treated patients either developed invasive fungal disease or died within 3 months (p=0007). A comparison of evaluable patients (15 in cyclophosphamide group, 14 in ATG group) at the 3month follow-up visit showed a significantly superior early response rate (all response categories) in the ATG group (11 [78%] vs seven [47%] responding; p=0035; figure 3). The response rate at 6 months was still higher for the ATG group but not significantly so (nine of 12 [75%] vs six of 13 [46%], p=010). The quality of responses is also shown in figure 3. Three of four eligible patients assigned cyclophosphamide have crossed over to treatment with ATG; however, neither of two eligible patients assigned ATG have crossed over to cyclophosphamide. At median follow-up of 667 days, four deaths had occurred (three in the cyclophosphamide group as detailed above, one in the ATG group due to refractory pseudomonal sepsis on day 151); however, the overall survival comparison did not reach significance (p=026).

Discussion
Laboratory studies have suggested that most communityacquired aplastic anaemia is the result of autoimmune destruction of haemopoietic cells,13 and various immunosuppressive regimens are effective in improving marrow function and increasing peripheral blood counts.1 Cyclophosphamide is a potent depressant of immune function, at high doses resulting in sustained decreases in both number and function of T and B cells.14,15 It is effective therapy in autoimmune diseases,1618 and as a single agent is an adequate conditioning agent for bonemarrow transplantation, sufficiently immunosuppressive to allow full donor engraftment. Johns Hopkins University Hospital has systematically

applied high-dose cyclophosphamide as first treatment in severe aplastic anaemia, having now reported on 24 patients.11,19 The haematological response rate is similar to that generally reported for ATG, but relapses and evolution to clonal haematological disease did not occur, and the blood counts of many of the treated patients were normal. In our randomised trial, some patients treated with cyclophosphamide showed excellent haematological responses, consistent with this published experience. Nevertheless, although the period of post-treatment evaluation in our series is short, relapses have already occurred in both study groups (one in the cyclophosphamide group, five in the ATG group) yet all were responsive to reinstitution of immunosuppression. Relapsed aplastic anaemia after syngeneic transplantation and conditioning with high-dose cyclophosphamide has also been reported.20 In our series, three patients who presented with one clonal complication of aplastic anaemia, cell-surface glycosylphosphoinositol-anchored protein deficiency, which is diagnostic of paroxysmal nocturnal haemoglobinuria, did not show any change in this abnormal cell population after cyclophosphamide treatment. A chromosomal abnormality classically attributed to myelodysplastic syndrome has been detected in one complete responder in the ATG group (trisomy 8); however, detection of this abnormality 3 months after treatment raises the possibility that it was present but not detectable due to poor in-vitro marrow-cell growth at presentation. The overlap between aplastic anaemia and clonal haematological disorders may indicate that these disorders represent different expressions of a common pathobiology and not its treatment.2123 The reason for early termination of our trial was the high toxicity of cyclophosphamide in these patients. Neither fungal disease nor early death occurred among patients who were assigned ATG in this randomised trial, and the single ATG-group patient who died (beyond 5 months) presented with complete absence of neutrophils and showed no improvement despite repeated courses of ATG and administration of haemopoietic growth factors. Such patients with extreme neutropenia have a notoriously poor prognosis.24 In our previous experience of 122 patients treated with ATG plus ciclosporin, those presenting with absolute neutrophil counts below 020109/L had mortality at 3 months of 28%; the early course of patients with neutrophil counts above that value was much better, with only two of 75 dying within 3 months of treatment (unpublished). Among the 15 patients treated with cyclophosphamide and ciclosporin in the trial reported here, three died within a short period of initial therapy, and a further three had very serious, life-threatening fungal infections that would probably have been fatal had the patients not been aggressively supported with granulocyte transfusions until they experienced some recovery of their own neutrophil production. Only one of these six patients presented with super-severe disease. Two other patients, who were not part of the randomised protocol, were treated with high-dose cyclophosphamide by compassionate exemption (previous treatment with ATG in one, the presence of severe paroxysmal nocturnal haemoglobinuria in the other) and died early of infectious complications (disseminated fusariosis and bacterial sepsis). Even among the Johns Hopkins patients, early deaths occurred in two of ten in the pilot series and in three of 14 in the later group; mortality was secondary to sepsis in one, fungal infection in two, and haemorrhage in two.19 High-dose cyclophosphamide has been used to treat severe aplastic anaemia in less developed countries, where it may be favoured over ATG because it is cheaper

THE LANCET Vol 356 November 4, 2000

1557

ARTICLES

and simpler to administer. However, reports from Mexico,25 Poland (R Rokicka-Mileska, personal communication), and Korea (Hyeoung Joon Kim, personal communication) indicate high mortality. Persistent neutropenia is a key variable predisposing immunocompromised patients to invasive fungal infections, particularly aspergillosis.26 In patients treated for acute leukaemia, the major risk factor for pulmonary aspergillosis is duration of neutropenia, with a progressive rise in cumulative frequency of infection to more than 70% after 34 days or longer of neutropenia.27 The patients in our study who received cyclophosphamide sustained a significantly greater duration of profound neutropenia than did those receiving ATG, with a median time to reach an absolute neutrophil count of 050109/L of 53 days versus only 3 days. We believe that this key risk factor substantially increased the risk of invasive fungal infections in patients receiving cyclophoshamide. That the mycoses in these patients were due principally to Aspergillus species is further consistent with the role of profound and persistent neutropenia as the major risk factor. Because a proportion of patients in the original cyclophosphamide series received ciclosporin,9 patients in both groups of our study received this drug. A contribution of this immunosuppressive agent to the higher frequency of invasive fungal disease in the cyclophosphamide group is possible but unlikely. Host defences against Aspergillus spp depend on the oxidative metabolism of neutrophils and monocytes. In-vitro studies show that ciclosporin can induce modest suppression of superoxide production and hyphal damage by granulocytes at high concentrations, but these changes are incremental and not observed as a net effect of immunosuppression in vivo.28 Furthermore, ciclosporin does not substantially increase the direct risk of invasive aspergillosis in non-neutropenic animal models.29 Even if these ciclosporin-induced in-vitro effects had an in-vivo correlate, fungal infections in our patients occurred during the long period when white-blood-cell counts were barely measurable, thus abrogating any potential direct effect on these phagocytic cells. The absence of invasive aspergillosis in patients assigned ATG plus ciclosporin is consistent with these experimental data. Finally, several studies have suggested that ciclosporin alone may have potentially beneficial direct antifungal effects.30 Thus, the simplest explanation for the difference in the early clinical course between cyclophosphamide and ATG is that the cytotoxic action of cyclophosphamide means that patients with good neutrophil numbers become at high risk, in the category of super-severe disease. In undertaking this trial, we mistakenly expected that patients with higher neutrophil counts would recover from cyclophosphamide-induced pancytopenia with kinetics similar to those in patients given equivalent doses for other disorders. However, the recuperation was delayed. By contrast, neutropenia after ATG was brief, and with the addition of G-CSF, some patients showed very rapid increases. Although not myeloablative, high-dose cyclophosphamide destroys late haemopoietic progenitor cells,31,32 which may be important in maintaining peripheral blood counts in aplastic anaemia. The toxic effects of cyclophosphamide on residual stem cells and the mutagenic effects of an alkylator do not seem to have an adverse effect in the long term in the treated patients so far reported, but these observations require confirmation. Whether such a study can be conducted ethically is questionable. Stem-cell transplantation, an accepted option for the treatment of aplastic anaemia, also carries acute risks, but the benefits of that procedure have been

amply documented, and more importantly, patients now can be logically guided by age and severity through appropriate evidence-based algorithms.4 In our experience, cyclophosphamide toxicity was not predictable from age or disease status, and therefore patients cannot be rationally selected. With improved survival recently reported for conventional regimens of ATG and ciclosporin,33 the potential usefulness of adjunctive haemopoietic growth factors, and the advent of new classes of potent specific immunosuppessive drugs (mycophenolate mofetil, rapamycin, anti-interleukin-2 receptor monoclonal antibody, and others), high-dose cyclophosphamide seems to be a dangerous choice for the treatment of aplastic anaemia patients with many other therapeutic options and an increasingly improving prognosis.33
Contributors
John Tisdale was responsible for protocol design, writing, submission, screening of patients, follow-up, termination, and report preparation. Daniel Dunn contributed to protocol design and writing, early screening of patients, and consent. Nancy Geller designed the statistical section, shared in planning, endpoint definitions, and analysed the trial at termination. Michelle Plant arranged randomisation, followed up all inpatients daily, monitored drug compliance, educated patients on all study medications, and retrieved drug-related data. Olga Nunez arranged screening and admission and follow-up of all study patients, and collected and recorded all study data. Cynthia Dunbar was involved in Hematology Branch protocols, was training programme chief, and supervised all fellows involved in the protocol. John Barrett was involved in Hematology Branch protocols, and was Chief of the Stem Cell Transplant Unit. Thomas Walsh cared for patients with proven or suspected fungal disease, and contributed to the other care of patients and preparation of the report section on fungal complications. Stephen Rosenfeld designed the database for patients on trial, and was involved in Hematology Branch protocols. Neal Young, Hematology Branch Chief, established our large referral centre for aplastic anaemia, and oversaw design and implementation and initiated the current trial.

Acknowledgments
We thank the staff of 2W, OP7, 13E, and 10D, for excellent care of patients.

References
Young NS, Barrett AJ. The treatment of severe acquired aplastic anemia. Blood 1995; 85: 336777. 2 Frickhofen N, Kaltwasser JP, Schrezenmeier H, et al. Treatment of aplastic anemia with antilymphocyte globulin and methylprednisolone with or without cyclosporine: the German Aplastic Anemia Study Group. N Engl J Med 1991; 324: 1297304. 3 Rosenfeld SJ, Kimball J. Vining D, Young NS. Intensive immunosuppression with antithymocyte globulin and cylosporine as treatment for severe aplastic anemia. Blood 1995; 85: 305865. 4 Bacigalupo A, Brand R, Oneto R, et al. Treatment of acquired severe aplastic anemia: bone marrow transplantation compared with immunosuppressive therapy: the European Group for Blood and Marrow Transplantation experience. Semin Hematol 2000; 37: 6980. 5 Schrezenmeier H, Marin P, Raghavachar A, et al. Relapse of aplastic anaemia after immunosuppressive treatment: a report from the European Bone Marrow Transplantation Group SAA Working Party. Br J Haematol 1993; 85: 37177. 6 Socie G, RosenfeldS, Frickhofen N, Gluckman E, Tichelli A. Late clonal diseases of treated aplastic anemia. Semin Hematol 2000; 37: 91101. 7 Dunn DE, Tannawattanacharoen P, Boccuni P, et al. Paroxysmal nocturnal hemoglobinuria cells in patients with bone marrow failure syndromes. Ann Intern Med 1999; 131: 40108. 8 Math G, Amiel JL, Schwarzenberg L, et al. Bone marrow graft in man after conditioning by antilymphocytic serum. BMJ 1970; 2: 13136. 9 Thomas ED, Storb R, Giblett ER, et al. Recovery from aplastic anemia following attempted marrow transplantation. Exp Hematol 1976; 4: 97102. 10 Baran DT, Griner PF, Klemperer MR. Recovery from aplastic anemia after treatment with cyclophosphamide. N Engl J Med 1976; 295: 152223. 11 Brodsky RA, Sensenbrenner LL, Jones RJ. Complete remission in severe aplastic anemia after high-dose cyclophosphamide without bone marrow transplantation. Blood 1996; 87: 49194. 1

1558

THE LANCET Vol 356 November 4, 2000

ARTICLES 12 Brodsky RA. Biology and management of acquired severe aplastic anemia. Curr Opin Oncol 1998; 10: 9599. 13 Young NS, Maciejewski J. The pathophysiology of acquired aplastic anemia. N Engl J Med 1997; 336: 136572. 14 Brinkman CJJ, Nillsen WM, Hommes OR. The effect of cyclophosphamide on T lymphocytes and T lymphocyte subsets in patients with chronic progressive multiple sclerosis. Acta Neurol Scand 1984; 69: 9096. 15 Mickey MR, Ellison GW, Fahey JL. Correlation of clinical and immunological states in multiple sclerosis. Arch Neurol 1987; 44: 37175. 16 DeVita S, Nero R, Bombardieri S. Cyclophosphamide pulses in the treatment of rheumatic diseases: an update. Clin Exp Rheumatol 1991; 9: 17993. 17 Valente RM, Conn DL, Current therapies for systemic vasculitis. Semin Neurol 1994; 14: 38086. 18 Fox DA, McCune WJ. Immunosuppressive drug therapy of systemic lupus erythematosus. Rheum Dis Clin North Am 1994; 20: 26599. 19 Brodsky RA, Smith BD, Brodsky I, Jones RJ. High-dose cyclophosphamide (Cy) without allogeneic bone marrow transplant (BMT) produces durable complete remissions in severe aplastic anemia (SAA). Blood 1999; 94: 674 (abstr). 20 Hinterberger W, Rowlings PA, Hinterberger-Fischer M, et al. Results of transplanting bone marrow from genetically identical twins into patients with aplastic anemia. Ann Intern Med 1997; 126: 11622. 21 Fuhrer M, Burdach S, Ebell W, et al. Relapse and clonal disease in children with aplastic anemia (AA) after immunosuppressive therapy (IST): the SAA 94 experience. Blin Paediatr 1998; 210: 17379. 22 Jin JY, Tooze JA, Marsh JCW, Matthey F, Gordon-Smith EC. Myelodysplasia following aplastic anaemia-paroxysmal nocturnal haemoglobinuria syndrome after treatment with immunosuppression and G-CSF: evidence for the emergence of a separate clone. Br J Haematol 1996; 94: 51012. 23 Molldrem J, Caples M, Mavroudis D, Plante M, Young NS, Barrett AJ. Antithymocyte globulin (ATG) abrogates cytopenias in patients with myelodysplastic syndrome. Br J Haematol 1997; 99: 699705. Locasciulli A, vant Veer L, Bacigalupo A, et al. Treatment with marrow transplantation or immunosuppression of childhood acquired severe aplastic anemia: a report from the EBMT SAA working party. Bone Marrow Transplant 1990; 6: 21117. Gmez-Morales E, Valle ES, Cabrera GC, Chvez JP. High dose cyclophosphamide without bone marrow transplantation in patients with severe aplastic anemia. Blood 2000; 94: 176 (abstr). Walsh TJ, Hiemenz J, Pizzo PA. Evolving risk factors for invasive fungal infections: all neutropenic patients are not the same. Clin Infect Dis 1994; 18: 79398. Gerson SL, Talbot GH, Hirwitz S, Strom BL, Lusk EJ, Cassileth PA. Prolonged granulocytopenia: the major risk factor for invasive pulmonary aspergillosis in patients with acute leukemia. Ann Intern Med 1984; 100: 34551. Roilindes E, Robinson T, Sein T, Pizzo PA, Walsh TJ. In vitro and in vivo effects of cyclosporin A on phagocytic host defenses against Aspergillus fumigatus. Antimicrob Agents Chemother 1994; 38: 288388. Berenguer J, Ali N, Allende MC, et al. Itraconazole in experimental pulmonary aspergillosis: comparison with amphotericin B, interaction with cyclosporin A, and correlation between therapeutic response and itraconazole plasma concentrations. Antimicrob Agents Chemother 1994; 38: 130308. Kim JH, Perfect JR. Infection and cyclosporin. Rev Infect Dis 1998; 11: 67790. Trainor KJ, Morley AA. Screening of cytotoxic drugs for residual bone marrow damage. J Natl Cancer Inst 1976; 57: 123739. Porcellini A, Manna A, Talevi N, et al. Effect of two cyclophosphamide derivatives on hemopoietic progenitor cells and pluripotential stem cells. Exp Hematol 1984; 12: 86366. Bacigalupo A, Bruno B, Saracco P, et al. Antilymphocyte globulin, cyclosporine, prednisolone, and granylocyte colony-stimulating factor for severe aplastic anemia: an update of the GITMO/EBMT on 100 patients. Blood 2000; 95: 193134.

24

25

26

27

28

29

30 31 32

33

THE LANCET Vol 356 November 4, 2000

1559