We performed a survey of the European Cooperative Group for Blood and Marrow Transplantation to analyze the outcome of 625 acute promyelocytic leukemia (APL) patients transplanted with auto- or allogeneic-hematopoietic stem cell transplantation (autoHSCT, alloHSCT) after 1993, in first (CR1) or in second complete remission (CR2). Leukemia-free survival (LFS) at 5 years in CR1 was 69% for 149 patients autografted and 68% for 144 patients allografted, whereas in CR2, LFS was 51% in 195 autoHSCT and 59% in 137 alloHSCT recipients, respectively. In the group of autoHSCT for CR1 (n=149), higher relapse incidence (RI) was associated with shorter time from diagnosis to transplant (<7.6 months); transplant-related mortality (TRM) was increased in older patients (>47 years), whereas for CR2, longer time from diagnosis to transplant (>18 months) was associated with increased LFS and decreased RI. In the alloHSCT group for CR1 (n=144), age (<33 years) was associated with increased LFS and decreased TRM and for CR2 (n=137), the use of mobilized peripheral blood stem cells was associated with decreased TRM. Female recipient, a female donor to male recipient and transplants performed before 1997 were associated with decreased RI. In conclusion, HSCT still appears to have a role in APL, especially for patients in CR2.
Several large multicenter studies1, 2, 3, 4, 5, 6, 7 have shown that regimens combining upfront all-trans retinoic acid (ATRA) and chemotherapy lead to a high curability rate in acute promyelocytic leukemia (APL) that currently exceeds 70% of cases, suggesting that hematopoietic stem cell transplantation (HSCT) need not be used to consolidate patients in first complete remission (CR1). In spite of this progress, however, treatment failure still occurs in approximately 10–25% of patients receiving state-of-the art therapy due to early death or, more frequently, disease relapse.
When clinical relapse occurs, or when it is predicted accurately by molecular monitoring,8 HSCT remains a widely adopted strategy as a part of the salvage therapy.9, 10, 11 However, there is no general consensus on the choice of transplant type, autologous (autoHSCT) or allogeneic (alloHSCT) in this setting. In this regard, HSCT results in the ATRA era, and particularly outcome data of auto and alloHSCT in CR2, have only been reported in small and non-comparative patient series.9, 10, 11 Results from these studies have suggested that autoHSCT in CR2 is still associated with a high probability of long-term survival, particularly for patients undergoing the procedure while in molecular remission (i.e. testing polymerase chain reaction (PCR)-negative for the PML/RARα (promyelocytic leukemia/retinoic acid receptor α) hybrid gene pre-transplant).9, 12 In a recent analysis of the European APL group, allogeneic HSCT for 23 APL relapsed patients was associated with higher transplant-related mortality (TRM) compared with autologous HSCT.13 Since the last EBMT survey by Mandelli et al.14 in 1994, no studies in large patient series have been reported and more importantly there are no recent data available on stem cell transplantation in CR1, which is considered as a minor, if not unreasonable therapeutic option. Therefore, it is of interest to analyze the current retrospective results and risk factors of HSCT in APL to investigate further the place of autoHSCT and alloHSCT in APL (both in CR1 and in CR2) after the advent of ATRA.
Patients and methods
Data for 625 adult APL patients undergoing HSCT were reported to the Acute Leukemia Working Party of the European Cooperative Group for Blood and Marrow Transplantation (EBMT). Diagnosis of APL was based on morphological criteria according to the French–American–British classification.15 Results of HSCT in patients with APL transplanted before 1993 were analyzed partially by Mandelli et al.14 and reported in 1994. In the present study, we analyzed the results of HSCT in 625 patients in CR (293 in CR1 and 332 in CR2) transplanted between January 1993 and December 2003. In the CR1 group, 149 and 144 patients received an autoHSCT and an alloHSCT, respectively, whereas 195 and 137 patients in CR2 were autografted and allografted, respectively. For the alloHSCT group, only patients with a graft from an HLA-matched sibling were included in the study. The main clinical characteristics of the patients are reported in Table 1.
Owing to the type of study, consisting of a retrospective analysis based on information obtained from the registry, no data were available as to whether or not patients had received ATRA before stem cell transplantation (SCT). Given the considered historic period (1993–2003), however, we assume that most patients in study had been treated with ATRA and chemotherapy, as this was the strategy adopted commonly in most European hematology institutions. Similarly, no information on patient molecular status (reverse transcriptase PCR of PML/RARα) pre- and post-SCT was available.
End points definitions and statistical analysis
Data were analyzed as of 30 June 2005. Median follow-up was 42 months (range 1–132 months).
Four outcomes were studied in this series: (i) TRM was defined as all causes of non-leukemic deaths; (ii) relapse incidence (RI) was defined on the basis of morphological evidence of leukemia in bone marrow, or other extramedullary organs. To evaluate the probability of relapse, patients dying either from direct toxicity of the procedure or from any other cause not related to leukemia were censored; (iii) leukemia-free survival (LFS) was defined as time interval from transplant to first event (either relapse or death in CR); and (iv) overall survival (OS). All outcomes were calculated for CR1 and CR2 patients separately.
Patient or graft characteristics (listed in Tables 2 and 3) were analyzed for their potential prognostic value on each of the outcomes and in each type of transplant (autoHSCT or alloHSCT). For continuous variables, we used either clinically meaningful cut points (interval diagnosis to CR2 less or more than 18 months) or the median as a cut point. Statistical analyses were performed independently for each end point. The incidence of each event was non-parametrically estimated. Probability of OS and LFS were estimated by the product-limit method.16 The significance of differences between curves was estimated by the log-rank test (Mantel–Cox). Then, all variables were included in Cox proportional hazard model.17 To have a minimum follow-up period of 18 months, the analysis was performed in June 2005 on patients transplanted until December 2003.
Relapse and non-relapse mortality are mutually competing events. Accordingly, estimations of incidence of these events relied on the non-parametric estimator of cumulative incidence curves, whereas multivariate analyses were based on the proportional hazard model for this subdistribution of competing risks.18 These analyses were performed using the cmprsk package (developed by Gray, June 2001) on Splus 2000 and SPSS softwares.
Number of HSCT in Europe from 1993 to 2003
Table 1 lists the number of HSCT (auto or allo) for patients in CR1 or CR2. The number of HSCT has decreased progressively for patients in CR1 since 1998; however for those patients transplanted in CR2, it has remained stable until 2002. In 2002 and 2003, there was also a decrease for patients transplanted in CR2.
Patients in CR1
The cumulative incidence of TRM and RI at 5 years were 10±3 and 21±4%, respectively. The 5-year estimate of LFS was 69±4% (Figure 1a). Table 2 shows the univariate analysis of variables associated with TRM, RI and LFS. In a multivariate analysis, higher RI was associated with shorter time from diagnosis to transplant (<7.6 months) (HR=0.32, 95 CI=0.14–0.75, P=0.009), and TRM was increased in older patients (>47 years) (HR=3.7, 95 CI=1.3–10.5, P=0.045) (Figure 1b). No risk factor was selected in the model for LFS.
Patients in CR2
For 195 patients transplanted in CR2, the 5-year cumulative incidence of TRM was 16±3% and the RI was 37±4%. LFS at 5 years was 51±4% (Figure 1c). In a univariate analysis, the 5-year estimates of RI and LFS were 31±4 and 56±5% for patients transplanted 18 months after diagnosis compared respectively with 25±7 and 53±6% for those transplanted earlier (Table 2). In a multivariate analysis, only longer time from diagnosis to HSCT (>18 months) was associated with increased LFS (HR=0.43, 95 CI=0.27–0.65, P<0.0001) and decreased RI (HR=0.42, 95 CI=0.25–0.71, P=0.001). No statistically significant risk factor was found to be associated with TRM.
Patients in CR1
Acute GVHD II–IV was observed in 30% of the patients (32 patients had grade II, eight grade III and five grade IV) and chronic GVHD was observed in 36% of patients at risk. The 5-year estimates of TRM, RI and LFS for patients were 20±3, 12±3 and 68±4% (Figure 2a), respectively. Table 3 lists the univariate analysis for 5-year outcomes.
The 5-year estimates of cumulative incidence of TRM and LFS for patients younger that 33 years were 10±4 (Figure 2b) and 78±5%, respectively. There was a trend of lower TRM for patients allotransplanted more than 6 months after diagnosis (Table 3). In a multivariate analysis, recipient age (>33 years) was associated with decreased LFS (HR=2.25, 95 CI=1.20–4.20, P=0.01) and higher TRM (HR=3.32, 95 CI=1.38–8.00, P=0.007). A longer time from diagnosis to HSCT (>6 months) was also an independent factor associated with lower TRM (HR=0.46, 95 CI=0.22–0.95, P=0.036). No risk factor studied was associated with the RI in this setting.
Patients in CR2
Acute GVHD II–IV was observed in 35% of patients (28 patients had grade II, 10 grade III and 11 grade IV) and chronic GVHD was observed in 39% of patients at risk. The 5-year estimate of LFS was 59±4% (Figure 2c). Table 3 lists the univariate analysis of LFS, TRM and RI at 5 years.
Cumulative incidence of TRM at 5 years was 24±1%. In univariate analysis, cumulative incidence of TRM was 12±4% for allotransplants using peripheral blood stem cells (PBSC) and 31±4% for those using bone marrow cells as a source of hematopoietic stem cells (P=0.008). There was also significant statistical association of TRM with the period of transplant, that is 15±4% for patients transplanted after 1997 compared with 31±5% for those transplanted before this period (P=0.03). In a multivariate analysis for TRM, only PBSC transplants were associated with lower TRM (HR=0.30, 95 CI=0.12–0.77, P=0.03).
Cumulative incidence of relapse at 5 years was 17±1%. Results of univariate analysis are listed in Table 3. High white blood cells (WBC) (>median) at diagnosis were a statistically significant factor (P=0.008) associated with higher relapse rate (Table 3). In a multivariate analysis, the following factors were associated with decreased incidence of relapse: (i) transplants performed before 1997 (HR=4.8, 95 CI=1.8–13, P=0.002), (ii) female recipient (HR=0.30, 95 CI=0.10–0.84, P=0.02) and (iii) female donor to male recipient (HR=0.14, 95 CI=0.03–0.55, P=0.005). High WBC count was not included in the model due to absence of relapse for those patients with lower WBC count at diagnosis.
This survey of the EBMT Group shows that HSCT activity in Europe has decreased progressively since 1998 for patients with APL in CR1, whereas the number of patients transplanted in CR2 has remained stable over the study period. In addition, the data from this survey confirm a quite similar LFS in allo- and autoHSCT groups, as well as a lower TRM and a higher RI in autoHSCT than in alloHSCT. According to the type of transplant, we were also able to find certain risk factors for outcomes such as age, time from diagnosis to transplant, year of transplants, source of stem cells and donor gender.
The first objective of our study was to perform a survey on the results of HSCT for APL patients in the ATRA era. To our knowledge, no studies including large series of patients with APL receiving an auto- or alloHSCT have been reported after the first EBMT survey published in 1994 by Mandelli et al.14 when ATRA was not yet used in the majority of the European centers. When we planned the present study, we did not expect to discover such a high number of patients with APL still transplanted after 1993, a number equivalent to those transplanted before 1993 (over a similar period of time), which at that time included mostly patients treated with only chemotherapy.
Although we were unfortunately unable to obtain complete data relative to pre-transplant treatments in the registry, it is conceivable that most patients here analyzed had received modern ATRA-containing regimens pre-HSCT. In fact, the present study included patients transplanted after 1993, when front-line ATRA therapy was adopted for both newly diagnosed and relapsed APL in most European Centers. However, we have observed that the number of patients transplanted in CR1 with an auto- or alloHSCT has decreased progressively since 1998, whereas the number of patients transplanted in CR2 has remained quite stable but with a decrease in 2003, probably reflecting the adoption for these patients of other therapeutic strategies such as arsenic trioxide (ATO). Although the justification for transplanting patients in CR2 is quite obvious, it is unclear why patients with APL were transplanted in CR1. However, those transplants were performed mostly during the earlier years after the introduction of ATRA-based regimens, whereas only few patients in CR1 are currently transplanted, in particular those showing persistent molecular disease after front-line consolidation.
Compared with the previous EBMT survey,14 the present EBMT study shows a higher LFS, confirming a quite similar final outcome in allo- and autoHSCT groups. This similar final outcome seems related to a significant reduction of TRM in the last years (8% in autoHSCT and 17% in alloHSCT) and to the counterbalance between the higher RI and TRM rates observed in the auto- and alloHSCT groups, respectively. In addition to some improvement over time, these better results may be related to the wide use of ATRA pre- and/or post-transplant.
Our second objective was to analyze risk factors for outcomes, mainly TRM, RI and LFS for patients receiving either auto- or alloHSCT in CR1 and most importantly in CR2. As expected, in first CR, patient's age, that is younger than 33-years old, was the most important factor associated with decreased TRM in both auto- and alloHSCT settings. Therefore, the option of alloHSCT should be considered seriously for younger patients in CR1 with persistent molecular disease after consolidation, whereas for older patients who test PCR-positive at this time point, other treatment options, such as ATO or gemtuzumab-ozogamicin may be considered.
In CR2, both approaches also produced reasonably good results. In the autoHSCT setting, LFS at 5 years was 51% for the entire study period and more than 60% after 1997. Importantly, patients with longer CR1 duration (diagnosis to transplant >18 months) had a better LFS and a lower RI. This observation makes sense because patients relapsing early are more likely to having received autografts still contaminated by tumor cells. Indeed autografting is recommended presently in APL patients in CR2 only if disappearance of minimal residual disease by PCR has been documented.19 Nevertheless, some investigators have proposed that the role of autoHSCT to consolidate high-risk patients (i.e. with hyperleucocytosis) should be investigated.19 Unfortunately, in our retrospective study, which was based on data from a registry, no information was available about the PML/RARα molecular status of the graft and/or of patient bone marrow at transplant. In addition, other relevant information was neither available, including the type of previous front-line nor salvage therapy used, the reason for the indication of transplant in CR1, as well as the type of transplant, among others.
For APL patients with an HLA-identical sibling, alloHSCT continues to be recommended in most centers as the treatment of choice for patients in CR2. We have found that the use of PBSC was associated with decreased TRM as compared to bone marrow. This finding is in agreement with the results of randomized studies comparing bone marrow and PBSC transplants, which showed better results for PBSC in patients transplanted with more advanced disease.20, 21, 22, 23, 24, 25, 26, 27 Interestingly, decreased RI was associated with a female donor to male recipient suggesting a graft-versus-leukemia effect mediated probably by minor H-Y antigens.28 Of note, we have observed that a higher WBC count at diagnosis was also an important risk factor for increased RI after allotransplantation. Another interesting finding is that post-transplant RI has increased after 1997, probably reflecting a selection of patients with more aggressive disease relapsing after receiving state-of-the-art front-line regimens.
In conclusion, the present study indicates that HSCT has been decreasing over the years for patients with APL in CR1 in Europe, but it has continued to be part of the treatment strategy for patients in CR2. The data from this survey indicates that, in spite of the favorable results with both auto- and alloHSCT (LFS around 70% at 5 years), HSCT should no longer be used to consolidate CR1 in patients with APL treated with modern ATRA plus chemotherapy regimens. Only selected and young patients with persistent molecular disease can probably benefit from this approach. Likewise, our results show that a high proportion of patients in CR2 achieve long-term OS after auto- and alloHSCT, and both procedures represent valid therapeutic options in this setting. The choice of one or other procedure will depend on the availability of an HLA identical donor and the time from diagnosis to transplant. Finally, in the near future, HSCT results will need to be compared with long-term results of other therapeutic options for APL relapse such as ATO and gemtuzumab ozogamicin, which have provided promising outcome data in small series with limited follow-up.
About this article
British Journal of Haematology (2016)