Late complications following treatment for severe aplastic anemia (SAA) with high-dose cyclophosphamide (Cy): follow-up of a randomized trial

Immunosuppression with antithymocyte globulin (ATG) and cyclosporine (CSA) results in long-term survival in patients with severe aplastic anemia (SAA), comparable to that achieved by allogeneic bone marrow transplantation from a histocompatible sibling donor,1 yet problems, including incomplete hematologic recovery, relapse, and the appearance of clonal hematopoietic disorders, complicate long-term management.2 High-dose cyclophosphamide has been proposed as an alternative immunosuppressive agent for the treatment of aplastic anemia, based on encouraging results in 2 uncontrolled trials3,4: in contrast to the experience with antithymocyte globulin (ATG)–based treatments, neither relapse nor clonal disease were reported. We initiated a phase III randomized trial to compare response rates to immunosuppression with either ATG or high-dose cyclophosphamide (Cy), both combined with CSA. Secondary endpoints included relapse and clonal evolution. Although primary response rates were not significantly different at 6 months, our trial was terminated early because of excess morbidity and early mortality in the Cy arm.5 We now report analysis of secondary endpoints after extended follow-up at a median of 38 months.

The protocol was approved by the Institutional Scientific Review Committee and the Institutional Review Board of the National Heart, Lung, and Blood Institute, and all patients gave written informed consent. A sample size of 91 patients per treatment arm was planned to allow comparison of the response proportions conducted at the 0.05 significance level, but the trial was terminated after accrual of only 31 patients.5 Secondary endpoints included relapse, the appearance of clonal hematologic disorders, and overall and event-free survival. All endpoints were assessed at scheduled follow-up visits at 6 months, 12 months, and yearly thereafter. Patients were considered responders if they experienced an improvement in blood counts sufficient to no longer meet criteria for severe disease, criteria that correlate with eventual transfusion independence.6 To better assess the quality of hematologic recovery, response was further classified with ordered, mutually exclusive criteria as partial response with transfusion dependence (PRd), partial response with transfusion independence (PRi), and complete response (CR; normal or near normal blood counts).5 Patients meeting criteria for sustained response of more than 3 months who subsequently experienced a fall in counts sufficient to require reinstitution of immunosuppressive drugs were considered to have relapsed. Peripheral blood samples were analyzed by flow cytometry for the presence of glycophosphatidyl inositol (GPI)–anchored protein-deficient granulocytes at presentation and at each scheduled follow-up; detection at 1.0% or greater on 2 or more evaluations was considered evidence for expansion of a clonal population of paroxysmal nocturnal hemoglobinuria cells (PNH). Evolution to myelodysplastic syndrome (MDS) was defined by the characteristic marrow morphology or the presence of a consistent chromosomal abnormality; cytogenetic examination of bone marrow samples was performed at presentation, at 6-month follow-up, and yearly thereafter.

Thirteen (81%) of 16 patients randomly assigned to ATG and 8 (53%) of 15 patients randomly assigned to Cy showed a hematologic response (Table 1). All responding patients, regardless of treatment allocation, eventually achieved transfusion independence (no remaining PRds), as predicted from previous results with standard immunosuppressive therapy.6Complete responses were observed in 10 (63%) of 16 patients in the ATG arm and 6 (40%) of 15 patients in the Cy arm (10 [77%] of 13 responders ATG, 6 [75%] of 8 responders Cy); there was no difference in the overall or complete response rates between arms (P = .12, P = .15, respectively, Fisher exact test).

Relapse is the most frequent long-term complication following immunosuppression with ATG-containing regimens,6,7 but it was not observed in the early Cy-treated patients.3Indeed, relapse occurred in 6 (46%) of 13 responders in the ATG arm, but relapse was also observed in 2 (25%) of 8 responders in the Cy arm (P = .38, Fisher exact test). Four of 6 patients in the ATG arm and 1 of 2 patients in the Cy arm were retreated, based solely on a fall in the platelet count. Relapse in which blood counts again met criteria for severe disease was observed in 1 patient in the ATG arm and 1 patient in the Cy arm. Although the patient in the Cy arm presented initially with counts just satisfying severity criteria, relapse to supersevere disease occurred more than 2 years after attaining a CR. All relapsed patients responded to reinstitution of immunosuppression either with CSA or with ATG and CSA; 3 patients in the ATG arm and 1 patient in the Cy arm required the addition of ATG.

The association between PNH and aplastic anemia is well established; however, early studies suggesting evolution to PNH used the now outdated Ham test.8 Patients with clinical PNH were excluded from our randomized trial; however, flow cytometric methods now allow more sensitive detection of such clones in not only erythroid but also myeloid cells, increasing both sensitivity and specificity, and the use of such techniques argues that PNH is a common and early event in SAA.9,10 The simultaneous absence of 2 GPI-anchored proteins highly expressed on normal granulocytes, CD66b and CD16, at levels above our threshold of 1.0% was used as criteria for establishing evidence of PNH clonal expansion11; such an expanded population was detected in 5 patients in the ATG arm and 4 patients in the Cy arm, at presentation. A GPI-anchored protein-deficient population just below our cutoff was detectable at diagnosis in one additional patient in the ATG arm; one subsequent determination was just above the cutoff. Regardless of treatment allocation, the overall percentage of GPI-anchored protein-deficient granulocytes has not changed substantially over extended follow-up (Table 2). Further, treatment of a single patient with clinical PNH with high-dose Cy by compassionate exemption produced neither hematologic improvement nor a change in the percentage of GPI-anchored protein-deficient granulocytes over time. Although detection of the PNH phenotype among blood cells of patients with aplastic anemia is relatively common, clinical PNH is less frequently observed, and only one patient in each treatment arm subsequently has developed evidence of intravascular hemolysis.

The late occurrence of MDS is the most dire complication observed in patients with SAA and occurs in both successfully treated and persistently cytopenic patients.12 Evolution to myelodysplasia was not described in the 2 pilot series published to date.3,4 In the current protocol, marrow cytogenetic evaluation revealed the presence of abnormalities characteristic of MDS not only in 2 patients in the ATG arm (trisomy 8, 9 of 20 metaphases at 6-month follow-up, increasing to 19 of 20 at 4-year follow-up; and 20q−, 9 of 50 metaphases at 3-year follow-up with 0 of 6 at 4-year follow-up) but also in 1 patient in the Cy arm (trisomy 8, 4 of 20 metaphases at 1-year follow-up; P = .70, Fisher exact test) despite a normal karyotype in all cases at presentation. These cytogenetic abnormalities were noted, however, on routine marrow samples among patients who remain in clinical remission and underscore the importance of marrow sampling even among patients with stable, recovered blood counts. Further, such chromosomal abnormalities do not necessarily portend a poor prognosis in patients with SAA.13 14 

Although the early termination of our trial because of toxicity on the Cy arm does not permit a comparison of either response rates or long-term complications with adequate statistical power, the tempered enthusiasm for this approach we experienced as a result of the high degree of early toxicity is further dampened by our recent observations. Cy treatment does not prevent the familiar long-term complications experienced by patients treated with conventional immunosuppression for SAA, and, as such, the resulting early morbidity and mortality cannot be justified by their anticipated absence. The continued development of alternative immunosuppressive regimens that take into account both the late complications as well as early safety is thus warranted.