Customize your JAMA Network experience by selecting one or more topics from the list below.
Rosenfeld S, Follmann D, Nunez O, Young NS. Antithymocyte Globulin and Cyclosporine for Severe Aplastic Anemia: Association Between Hematologic Response and Long-term Outcome. JAMA. 2003;289(9):1130–1135. doi:10.1001/jama.289.9.1130
Author Affiliations: Hematology Branch (Drs Rosenfeld and Young and Ms Nunez) and Office of Biostatistics Research (Dr Follmann), National Heart, Lung, and Blood Institute, Bethesda, Md.
Context In most patients, aplastic anemia results from T-cell–mediated
immune destruction of bone marrow. Aplastic anemia can be effectively treated
by stem cell transplantation or immunosuppression.
Objective To assess long-term outcomes after immunosuppressive therapy.
Design, Setting, and Patients Cohort of 122 patients (31 were ≤18 years and 91 were >18 years)
with severe aplastic anemia, as determined by bone marrow cellularity and
blood cell count criteria, were enrolled in a single-arm interventional research
protocol from 1991 to 1998 at a federal government research hospital.
Interventions A dose of 40 mg/kg per day of antithymocyte globulin administered for
4 days, 10 to 12 mg/kg per day of cyclosporine for 6 months (adjusted for
blood levels), and a short course of corticosteroids (1 mg/d of methylprednisolone
for about 2 weeks).
Main Outcome Measures Survival, improvement of pancytopenia and transfusion-independence,
relapse, and evolution to other hematologic diseases.
Results Response rates were 60% at 3 months after initiation of treatment, 61%
at 6 months, and 58% at 1 year. The blood cell counts of patients who responded
no longer satisfied severity criteria and they were transfusion-independent.
Overall actuarial survival at 7 years was 55%. Survival was associated with
early satisfaction of response criteria (86% vs 40% at 5 years; P<.001) and by blood counts at 3 months (reticulocyte count or platelet
count of >50 × 103/µL predicted survival at 5 years
of 90% [64/71] vs 42% [12/34] for patients with less robust recovery [P<.001 by log-rank test]). There were no deaths among
responders more than 3 years after treatment. Relapse was common, but severe
pancytopenia usually did not recur. Relapse did not influence survival. Thirteen
patients showed evolution to other hematologic diseases, including monosomy
Conclusions Approximately half of patients with severe aplastic anemia treated with
antithymocyte globulin and cyclosporine have durable recovery and excellent
long-term survival. These outcomes were related to the quality of hematologic
In patients with aplastic anemia, bone marrow failure leads to pancytopenia;
death occurs secondary to infection, bleeding, or complications of severe
anemia.1 Clinical observations and laboratory
experiments have implicated an immune pathophysiological effect, in which
T cells destroy hematopoietic stem and progenitor cells.2 Allogeneic
bone marrow transplantation replaces stem cells. Current transplant methods
produce high cure rates, but only a minority of patients have histocompatible
siblings, and graft vs host disease is a frequent complication in older recipients.3
Observation of recovery despite failed bone marrow transplant led to
the treatment of aplastic anemia with immunosuppressive therapy alone. Antithymocyte
globulin improved blood cell counts in about half of treated patients.4,5 Cyclosporine improved outcomes of many
who had not responded to antithymocyte globulin.6,7 The
combination of antithymocyte globulin, which lyses lymphocytes, and cyclosporine,
which blocks T-cell function, has led to survival rates comparable with those
observed with transplant recipients.8-10
However, immunosuppression is an imperfect treatment. About a third
of patients fail to respond, and even responders often have chronically low
blood cell counts. Late complications include relapse of pancytopenia and
development of secondary clonal hematologic diseases like myelodysplasia.11 These problems have resulted in pessimistic commentary
suggesting that immunosuppression only "postpones the inevitable"12 use of more toxic therapies such as high-dose cyclophosphamide13 and uncertainty in the timing of risky alternative
donor transplants in children14 and matched
unrelated transplants in adults.15
We analyzed long-term outcomes of one of the initial protocols establishing
the use of combined immunosuppressive therapy in aplastic anemia to determine
the association between relapse and survival, and to determine the rate and
significance of evolutionary events.9
Patients referred to the Warren Grant Magnusson Clinical Center with
a diagnosis of severe aplastic anemia between December 1989 and January 1998
were evaluated for entry into the protocol. Referrals were from hematologists
in private practice or at academic institutions throughout the United States
and world. Data were collected until February 2002 and censored from the date
of last contact with the patient.
Study inclusion required bone marrow cellularity of less than 30% and
depression of at least 2 of 3 hematopoietic lineages: absolute neutrophil
count of 500/µL or less, absolute reticulocytes of 40 × 103/µL or less (≤60 × 103/µL after January
1993 to reflect changes in instrumentation), and platelet count of 20 ×
103/µL or less. Exclusion criteria included creatinine level
higher than 2 mg/dL (176.8 µmol/L), concurrent malignancy, recent radiation
or chemotherapy, and poor expectation of immediate survival. The protocol
was approved by the National Heart, Lung, and Blood Institute's institutional
review board and all patients or their parents gave informed consent.
Patients were treated at the Clinical Center. Immediate hypersensitivity
to antithymocyte globulin was assessed by skin test; individuals with a positive
reaction underwent desensitization. A dose of 40 mg/kg per day of equine antithymocyte
globulin (Upjohn, Kalamazoo, Mich) was administered intravenously on days
1 through 4. A dose of 1 mg/kg per day of methylprednisolone was given concurrently
to prevent or ameliorate serum sickness, usually for about 2 weeks. The initial
dose of cyclosporine was 12 mg/kg per day in adults and 15 mg/kg per day in
children aged 3 to 18 years, and adjusted to maintain serum levels of 200
to 400 ng/mL or for renal toxicity.
Supportive care was provided as previously described.9 After
September 1991, short courses of granulocyte colony-stimulating factor were
administered as clinically indicated, usually for evidence of infection such
as fever or localized inflammation in the setting of severe neutropenia. In
transfusion-dependent patients, the hemoglobin was maintained at a level higher
than 7 g/dL (>9 g/dL for patients with underlying cardiopulmonary disease).
Platelets were transfused prophylactically for blood levels lower than 10
× 103/µL and at higher levels in the setting of clinically
significant or symptomatic bleeding. Aerosolized pentamidine was administered
as prophylaxis against Pneumocystis carinii during
Bone marrow biopsy and aspiration, including cytogenetic study, were
performed before enrollment. Children and young adults had a chromosome assay
to rule out Fanconi anemia, and all were tested for paroxysmal nocturnal hemoglobinuria
(PNH) by the Ham test. Patients were hospitalized for the administration of
antithymocyte globulin and discharged when clinically stable—usually
at about 2 weeks. After discharge, patients returned to the care of their
physicians for laboratory tests weekly for the first month and biweekly thereafter.
Cyclosporine levels were measured every 2 weeks. Patients returned to the
Clinical Center for complete evaluation at 3 and 6 months and 1 year after
treatment and then annually. Bone marrow was examined morphologically and
cytogenetically at 6 months, 1 year, and then annually.
Response was defined as no longer meeting blood cell count criteria
for protocol inclusion in the absence of recent transfusions and granulocyte
colony-stimulating factor administration. We previously showed that this definition
correlated strongly with both independence from transfusion and survival at
1 year.9 Relapse was not defined by blood cell
counts but rather as any reinstitution of immunosuppressive therapy.
Binary group comparisons were made using the χ2 or Fisher
exact test as appropriate. Nonbinary group comparisons were made with the
Kendall s statistic. Survival curves were calculated
using the Kaplan-Meier method and compared using the log-rank test.16 Cumulative incidence curves were calculated as described.17 Cox regression was used to estimate hazard ratios
for both time-fixed and time-varying covariates. All significance tests were
2-sided. Calculations were performed using Mathematica (Wolfram Research,
Champaign, Ill) and SPLUS (Mathsoft, Seattle, Wash) statistical software packages.
We enrolled 122 patients. The ratio of male to female patients was 1.14:1
and the median age 35.0 years. Thirty-one of the patients were children or
adolescents younger than age 18 years—18 were younger than 12 years
and 4 were younger than 5 years. The absolute neutrophil count was less than
200/µL at presentation in 48 (39%) and less than 500/µL in 80
(66%) patients. In two thirds of cases, there was no apparent precipitating
event. Onset was associated with medications in 12 patients (10%), chemicals
in 10 (8%), posthepatitis in 12 (10%), Epstein-Barr virus infection in 1 (1%),
and a complication of pregnancy in 3 (3%). The median time to treatment after
diagnosis was 31 days. Median follow-up at the time of this analysis was 7.2
Response rates were 60% at 3 months after initiation of treatment, 61%
at 6 months, and 58% at 1 year. Five patients, all of whom had some early
blood cell count improvement, did not meet response criteria at 3 months but
were transfusion-independent and classified as responders by 6 months. Four
patients were transient responders at 3 months and no longer satisfied response
criteria by 6 months. On univariate analysis, the likelihood of response was
not associated with sex, delay between diagnosis and protocol entry, or pretreatment
absolute neutrophil count. However, younger age was associated with a higher
response rate (P = .04 with age treated as a continuous
variable). Virtually all patients who were classified as responders were transfusion-independent.
Overall survival of all patients at 7 years posttreatment was 55% (Figure 1). Patients younger than 50 years
had better survival than those older than 50 years (66% vs 38%; P = .01). Patients with absolute neutrophil counts of less than 200/µL
at presentation had worse survival than those with higher absolute neutrophil
counts (48% vs 58%; P = .02). Excess deaths in severely
neutropenic patients occurred mainly in the first few months after diagnosis.
There were 12 deaths in 48 patients with severe neutropenia within 3 months
of treatment compared with 3 deaths in 74 patients without severe neutropenia.
Sex, ethnicity, etiology, and duration of disease were not correlated with
Posttreatment status was strongly associated with long-term prognosis,
as defined by response criteria (blood cell counts inconsistent with severe
aplastic anemia, transfusion-independence) or by blood cell counts. For patients
classified as responders at 3 months, actuarial survival at 5 years was 86%
compared with 40% for nonresponders (P<.001; Figure 1B). For patients surviving to 3 months,
the median platelet and reticulocyte counts were each approximately 50 ×
103/µL. Survival of patients who had either or both counts
above the median value was 90% (64/71 alive) at 5 years compared with 42%
(12/34 alive) for patients with both counts below the median (Figure 1C). There were no deaths from disease more than 3 years
after treatment among patients classified as responders at 3 months (Figure 1B). For patients who responded qualitatively,
quantitative counts were still of prognostic importance. Survival for responders
with more robust blood cell count recovery was 91% at 5 years compared with
47% for responders with both counts below the median (P<.001; Figure 1D).
Sixteen patients (13%) died before their 3-month evaluation (mostly
due to fungal infection). Despite persistent thrombocytopenia, few patients
died of hemorrhage. Six patients died due to complications following bone
marrow transplant (immunosuppression failed in these patients). A total of
7 patients who entered the study ultimately underwent bone marrow transplantation
(all from matched, unrelated donors): 4 were nonresponders and died due to
transplant complications; 3 were patients classified as responders at 3 or
6 months who later evolved to myelodysplasia. Only 1 responder who relapsed
with a new finding of monosomy 7 survived after a matched unrelated transplant.
The requirement for further immunosuppressive therapy not specified
by the protocol was common, occurring in 26 of 74 patients classified as responders
at 3 months. The cumulative incidence of relapse among responders was approximately
35% at 5 years (Figure 2A). Age,
sex, delay between diagnosis and protocol entry, absolute neutrophil count
prior to treatment, or presumed etiology were not associated with relapse.
Qualitative blood cell counts at 3 months were not associated with relapse.
The impact of relapse on survival was assessed using a Cox regression
model with a time-varying covariate. At the point of relapse, a patient was
assumed to be at a fixed increased risk of death. All 3-month responders were
included in this analysis and relapse was not statistically significantly
associated with death (P>.10).
Our relapse definition was broad because many patients restarted cyclosporine,
but were only experiencing a modest decline in 1 or more blood counts. Recurrent
frank severe pancytopenia was unusual, and only a minority of patients required
retreatment with antithymocyte globulin. Of a total of 25 relapsing patients,
8 showed only declining blood cell counts and a further 6 again required transfusions;
only 1 patient failed to respond to reinstitution (or continuation in 1 case)
of cyclosporine, and none died. For 9 patients who had recurrent severe pancytopenia,
6 were retreated with antithymocyte globulin and 3 died.
While some relapsed patients died of the complications of severe pancytopenia,
most responded to the reintroduction of immunosuppressive therapy. However,
the majority of relapsed patients required continued administration of cyclosporine
after retreatment, usually at a low dose to maintain blood cell counts above
levels required to avoid transfusion. About 90% of relapsing patients continued
with cyclosporine therapy for 1 to 2 years after its reinstitution and about
60% were cyclosporine-dependent at 6 to 7 years.
Thirteen patients were considered to have evolved to a new hematologic
diagnosis (Figure 2B). Cytogenetic
abnormalities predominantly involved chromosome 7 (monosomy in 9 patients
and deletion 7p in 1 patient) and chromosome 8 (trisomy in 2 patients). Monosomy
7 usually occurred with either a minimal initial clinical response or clinical
relapse to severe pancytopenia. Seven patients with this finding died, 4 of
refractory pancytopenia and 3 after evolution to acute myelogenous leukemia.
In contrast, trisomy 8 was associated with an initial hematologic response
followed by dependence of blood cell counts on continued cyclosporine administration,
and a good long-term prognosis.
In both univariate and multivariate analysis of baseline factors, only
age was associated with evolution to a new hematologic diagnosis. Each decade
of life increased the risk of evolution by about 40% (P<.05). Overall, qualitative blood cell counts at 3 months were not
predictive of evolution using the simple risk stratification model (both platelet
and reticulocytes at 3 months <50 × 103/µL), but
evolution to monosomy 7 and its poor consequences were predicted by 3-month
platelet count alone (3.5-fold increased risk for patients with platelets
<50 × 103/µL; P = .02).
Survival after evolution was poor, with more than 60% of evolving patients—all
with monosomy 7—dead within 3 years of the event. When we assessed the
impact of evolution on survival using a Cox regression model with a time-varying
covariate, in which a patient was assumed to be at a different and fixed risk
of death at the point of evolution, evolved patients were at 7 times the risk
of death compared with those who had not evolved (P<.001).
Paroxysmal nocturnal hemoglobinuria (defined by a positive Ham test
result) developed in 7 patients within 6 months of diagnosis (in all but 1
case). Early onset was reflected in event curves, in which the risk of PNH
was 10% at 2 years and remained stable for at least 7 years (data not shown).
More extensive analysis of PNH was not undertaken because the Ham test was
supplanted by flow cytometric methods as this study was ending.
Historically, patients with severe aplastic anemia had a poor prognosis.
In the early 1970s, 80% to 90% of patients died of complications of pancytopenia
before 12 to 18 months had elapsed.18,19 The
introduction first of bone marrow transplant and later of antithymocyte globulin
regimens dramatically altered the clinical course of aplastic anemia. Seventy-five
percent of European patients undergoing either treatment in the 1990s could
expect a 5-year survival.20 Antithymocyte globulin
combined with cyclosporine has been especially effective in children,21 older adults who do not fare well with transplant,22 and in patients with severe neutropenia.23 Despite evident short-term efficacy, questions remain
about the long-term benefits of immunosuppression. Concerns have focused on
reported high rates of relapse, evolution to PNH and myelodysplasia, and leukemic
transformation. This study provides data about the long-term outcomes after
treatment of aplastic anemia with immunosuppression in a consistently treated
population, and the consequences to the patient with incomplete blood cell
count responses to the combined therapy of antithymocyte globulin and cyclosporine.
We defined relapse as the requirement for further immunosuppressive
therapy not specified by protocol on the conservative assumption that even
modestly declining blood cell counts presaged severe pancytopenia, but the
number of patients who developed frankly recurrent aplastic anemia was small.
In other studies, relapse has been variably defined. In a retrospective European
analysis, the actuarial risk of relapse (defined as a new requirement for
transfusion) was 35% at 10 years.24 In more
recent studies of antithymocyte globulin combined with cyclosporine, the relapse
rate was 9%, but the majority of responding patients required continued cyclosporine
therapy.23 These high relapse rates suggest
that a large proportion of patients may still be inadequately treated by combined
immunosuppressive therapy. However, relapse of aplastic anemia does not carry
the same dire prognosis as relapse of leukemia. Patients who experienced relapse
usually responded to reintroduction of immunosuppressive therapy, and we were
unable to demonstrate an effect of relapse on survival. Lack of association
between relapse and initial or recovery blood cell counts and occurrence of
relapse after full hematologic recovery further suggest a mechanism of persistent
or resurgent pathological immunity rather than stem cell exhaustion. Laboratory
studies have shown that molecularly defined inhibitory T-cell clones are reduced
but often persistent at low levels—even in recovered patients.25,26
Several hematologic diseases of stem cell clonal origin are associated
with aplastic anemia. Most prominent is PNH, which for many years was falsely
considered a common late development in treated patients, with rates of evolution
cited to be as high as 25%.27 However, flow
cytometric assays have allowed the identification of an abnormally expanded
clonal population of PNH cells in a large proportion of aplastic anemia patients
at presentation, and the development of a new clone after treatment is unusual.28
The second major clonal process, myelodysplasia, can be difficult to
distinguish clinically from aplastic anemia and may even respond to immunosuppressive
therapy.29 Myelodysplasia, which is recognizable
by a characteristic bone marrow morphology or by the detection of typical
cytogenetic abnormalities, appears to be a true late complication of treated
Clonal evolution after aplasia occurs in stereotypical patterns—most
frequently involving chromosomes 7 or 8. These aberrations have different
clinical consequences.30 Monosomy 7 appeared
as a new event, sometimes after multiple normal cytogenetic examinations and
was almost always associated with refractory relapse of pancytopenia or malignant
transformation. In contrast, trisomy 8 was relatively benign and consistent
with hematologic response, cyclosporine-dependence, and a stable clinical
Unexpectedly in this study, outcome was related to the presence of early
recovery and to the quality of the blood cell count response at 3 months after
receiving antithymocyte globulin. The flatness of the survival curve suggests
that a significant proportion, perhaps 25% of treated patients and 50% of
responding patients, are likely to be functionally cured and should not experience
major complications from aplastic anemia. For this identifiable group, immunosuppression
does more than postpone what was once considered inevitable. These patients
also appear to have the durable improvements credited to the more toxic use
of high-dose cyclophosphamide. Conversely, patients with suboptimal hematologic
recovery, even if transfusion-independent, remain at risk. How best to manage
the patients refractory to a single course of antithymocyte globulin remains
uncertain. Options include further immunosuppression31 and
stem cell transplant from donors other than histocompatible siblings, which,
while often successful in children, continues to carry significant risk of
Create a personal account or sign in to: