Risk factors for outcome after allogeneic stem cell transplantation in patients with advanced phase CML

Allogeneic hematopoietic stem-cell transplantation (HSCT) remains the only curative option for patients with advanced chronic myeloid leukemia (CML). However, outcome is dismal and of short follow-up. The objective of the study was to determine long-term outcome and risk factors in patients with a history of CML Blast Crisis (BC; n = 96) or accelerated phase (n = 51) transplanted between 1990 and 2018. At transplant, patients had a median age of 39 (range 7–76) years and were in ≥CP2 (n = 70), in AP (n = 40) or in BC (n = 37) with a diagnosis-HSCT interval of median 1.9 (range 0.3–24.4) years. Overall survival (OS) amounted 34% (95% CI 22–46) and progression-free survival (PFS) 26% (95% CI 16-36) at 15 years. Adverse risk factors for OS and PFS were low CD34+ count in the graft, donor age (>36 years) and BC. Cumulative incidence of Non-Relapse Mortality (NRM) was 28% (95% CI 18–38) and of relapse (RI) 43% (95% CI 33–53) at 15 years. PB-HSCT and HSCT after 2008 were favorable prognostic factors for NRM, while family donor and patient age >39 years were independently associated with higher RI. HSCT resulted in long-term OS in patients with advanced CML. OS was improved in non-BC patients, with donors ≤36 years and with higher CD34+ dose in the graft.


INTRODUCTION
The introduction of tyrosine kinase inhibitors (TKI) for BCR/ABL in patients with chronic phase (CP) chronic myeloid leukemia (CML) resulted for the first time in regular disappearance of the malignant clone, restoration of normal life expectancy, and even treatment free remission. In contrast, the outcome of patients with TKI resistance or of those who present with advanced disease is still dismal. In such cases, the median overall survival (OS) with first and further TKI generations is less than 12 months [1][2][3][4][5]. Patients presenting with de novo blast crisis (BC), a distinct entity as compared to the accelerated phase (AP) and BC developing during TKI treatment, had remarkably adverse results [6]. Despite promising short term results with Ponatinib in subgroup of patients with resistance to TKI or T315I mutations [7], the use of second and third generation TKI has done little to change the overall outcome of BC and AP. Independent prognostic factors for increased risk of death in the TKI era were myeloid immunophenotype, prior TKI treatment, patient age ≥58 years, LDH ≥ 1227 IU/ L, platelet count <102 GPT/l, no allogeneic stem cell transplantation (HSCT), blast phase from CP/AP and presence of chromosome 15 aberrations in patients with BC [1].
HSCT remains the only curative option in advanced phase, but data on outcome are scarce. Results in BC are clearly inferior to those of CP1 pointing to an unmet need for improvement [8]. For patients in CP after BC at HSCT, age ≥45 years, lower performance status (<80%), longer interval from BC diagnosis to HSCT (>12 months), myeloablative conditioning and unrelated HSCT were risk factors for inferior survival in a multicenter (n = 46) study [9].
The purpose of the study was to evaluate long-term results (over 15 years) on a large number of patients with advanced CML and analyze risk factors for outcome. Therefore, we analyzed outcome and risk factors in 147 patients with advanced disease (BC, AP and CP ≥ 2) transplanted in two centers with a follow-up period of up to 15 years.

PATIENTS AND METHODS
All patients with advanced disease (n = 147) transplanted between 1990 and October 2018 at the University Hospital of Hamburg Eppendorf, Germany and the Raisa Gorbacheva Memorial Institute for Children Hematology and Transplantology (RGMI), Saint Petersburg, Russian Federation were analyzed (Table 1). Clinical, cytogenetic, and molecular characteristics of CML including mutational status and additional gene mutations at diagnosis and at HSCT are provided in Table 1. HSCT was performed according to standard protocols approved by the Ethical Committee of the University Medical Center Hamburg, Eppendorf, Germany and The Raisa Gorbacheva Memorial Institute, RGMI, St. Petersburg, Russian Federation, and after written informed consent including consent for data collection and analysis. TKI was given to 40 patients because of MRD + (defined as BCR-ABL1/ ABL1 transcript ratio of <0.1% on two consecutive measurements) or prophylaxis, nine for hematological relapse, while 96 received no TKI post-HCT. Follow-up was performed at given intervals. Assessment of BCR-ABL/ABL transcript levels was done by quantitative real time (RT)-PCR at the departments laboratories according to the updated guidelines of the European LeukemiaNet (ELN) [10]. Pre-established donor/recipient-specific polymorphism were used in all patients post-HSCT for donor chimerism studies. Y-chromosome-specific sequences were determined in sexmismatched transplants by validated molecular quantitative RT-PCR techniques, which guarantee high sensitivity [10 -4 ] for the assessment of chimerism [11,12].

Study endpoints and definitions
Primary endpoints of this retrospective analysis were long-term overall survival (OS) calculated from the date of HSCT to death due to any cause. Secondary endpoints were progression free survival (PFS), incidence of NRM (date of HSCT to date of death in the absence of disease relapse) and RI (from date of HSCT to date of relapse). For all endpoints, patients alive were censored at the date of last contact. ELN criteria were applied for the definition of BC, AP, remission, and relapse [13]. Acute and chronic GvHD were graded and reported according to the standard clinical criteria. The first of three consecutive days with white blood cell count (WBC) > 1.0 gpt/L was considered as leucocyte engraftment and >20.000/µL platelets without transfusion as platelet engraftment. Primary graft failure was defined as no engraftment within 28 days and relapse according to reappearance of molecular, cytogenetic, and hematological disease characteristics as molecular, cytogenetic, and hematological relapse.

Statistics
Characteristics of patients were expressed as median and range for continuous variables and frequencies for categorical variables. Categorical

RESULTS
Patient characteristics are given in Table 1  Transplant characteristics, graft source, GvHD prophylaxis and post transplant TKI treatment is given in Table 1. Post-HSCT TKI was given for MRD + or relapse prophylaxis to 40, for treatment of relapse to 9 and 96 patients did not receive TKIs.
Of the 147 patients, 93.9% engrafted, one progressed immediately after HSCT and 5.4% had primary graft failure ( Table 2). WBC engraftment was observed a median of 16 (range 9-39) and platelet engraftment a median of 17 (range 6-63) days post-HSCT. With a median follow-up of 9 years, OS of all patients amounted to 38% (95% CI 30-46) at 10 years and 34% (95% CI 22-46) at 15 years, while PFS reached 30% (95% CI 22-38) at 10 years and 26% (95% CI 16-36) at 15 years ( Fig. 1). History of BC before HSCT did not influence OS, but phase at HSCT. Patients with BC independent of myeloid or lymphoid origin at HSCT had a trend for worse OS as compared to AP or ≥CP2 ( Fig. 2; p = 0.07, 41% vs 30% at 10 years) but OS in patients with AP at transplantation was not significantly different compared to those with ≥CP2. Interestingly, no difference was observed in patients with de novo BC as compared to the rest ( Figure S1).
The median time for molecular relapse was 5.3 (range 0.7-44.4) months. A total of 57 patients had hematological relapse, 17 despite receiving TKI and 40 without TKI (Table 2). TKI treatment positively influenced OS (p < 0.001; Figure S2). Hematological RI amounted to 43% (95% CI 33-53) at 15 years, median time to hematological relapse was 10.3 (range 1-61) months and plateauing after 5 years. Of the two groups with molecular or cytogenetic relapse the mortality rate was 50.0%, whereas for those with hematological relapse the mortality rate was 92.9%. DLI were given in 34 patients (Table 2). Patients receiving DLIs had a better OS than patients without DLI ( Figure S3; p < 0.001). The most frequent causes of death was relapse of the primary disease (57.6%). Fatality rate at relapse was related to the status at HSCT. A total of 32 (71.1%) out of the 45 patients with non-BC and 17 (89.5%) out of 19 patients with BC at HSCT died.

DISCUSSION
In this study, the long-term outcome of 147 patients after HSCT for advanced phase CML resulted in 34% at 15 years OS and 26% at   Table 3. Univariate analyses of risk factors for overall survival (OS), progression free survival (PFS), cumulative incidence of non-relapse mortality (NRM), and cumulative relapse incidence (RI).  15 years PFS. OS was influenced by stage of advanced disease at HSCT and was 36% after ≥CP2, 45% after AP, and 30% after BP at 14 years. In addition, we found that low CD34 + cell count/kg bw in the graft, donor age >36 years, and BC were unfavorable independent prognostic markers for OS and PFS. BM and HSCT before 2008 were independently associated with higher NRM, and related donor and patients age (>39 years) with higher RI.

Factors
The current results extend our previous evaluation [8] on advanced phase CML with twice the patient numbers and an observation period extended to 15 years. Previously published results had OS of 43% with a plateau at 3 years, while here OS was 41% at 5 years, 38% at 10 years, and 34% at 15 years. Similarly, PFS of 32% at 2 years is now 31% at 5 years, 30% at 10 years, and 26% at 15 years. There were no relapses 5 years after HSCT and RI has plateaued since, confirming the curative potential of HSCT in advanced phase in both transplant centers (superimposable outcome; Figure S4; p = 0.55). Our results compared favorably with those previously published for advanced phase CML showing 23.8% OS at 3 years despite being transplanted earlier in the course of the disease [9] and are similar to the results published by the German CML study group [14]. In our analysis, OS of patients with active BC at the time of HSCT amounted to 35.1% at 3 years, while the OS previously reported for patients with advanced phase (but not only in BC) was 51.3% at 3 years. Interestingly, OS in de novo BC was not different from BC, AP and ≥CP2.
Our results compare also favorably with the one of patients with TKI resistance or T315I mutation treated with ponatinib alone. Despite having short follow up, BC patients treated with ponatinib alone had inferior results than in our cohort and patients with AP (n = 18) had similar short term results in both studies, but missing long-term results (follow up in the two studies is 36 months versus 16 years in our study). Future studies investigating the promising concept of sequential combination Ponatinib and HSCT are warranted.
CD34 + cell count/kg bw in the graft has been shown to be predictive for survival in different haematological diseases following PB-HSCT (CD34 + > 4.5 × 10 6 /kg bw only 18 CML patients) and BM-HSCT (CD34 + > 3 × 10 6 /kg bw; 55 patients with CML, of which 11 accelerated phase and 2 in BC) [15,16]. CD34 + cell dose (>2.5 × 10 6 /kg bw) was found to influence OS in 99 AML patients transplanted with BM grafts from HLA identical siblings [17]. We found higher CD34 + dose to be associated with better OS and PFS. This correlation was especially significant in patients transplanted with BM. Related (RD) and unrelated donors (UD) may have influenced indirectly CD34 + cell dose following differences in donor age. While there was no difference in patients age [median 39  years and 39    for this effect may also involve the stage of the disease (CML in advanced phase) and the higher number of CD34 + (median 5.4 × 10 6 /kg bw) used in our cohort as compared to the previously published results. The multivariable analysis confirmed that the CD34 + dose and not the graft source influenced outcome. Furthermore, a multivariate analyses only on patients transplanted with BM grafts, which usually have less CD34 + cells, confirmed CD34 + cell dose as an independent variable (data not shown). Accellerated engraftment after higher CD34 + cell dose may decrease NRM and possibly lead to a lower RI (see univariable analysis) finally resulting in better outcome. The advantage of PB-HSCT over BM-HSCT has been a matter of debate for several decades and is associated indirectly with higher CD34 + cell dose. PB-HSCT has been shown to have faster neutrophil and platelet engraftment and less severe acute and extensive chronic GvHD [18]. Another publication reported that PB-HSCT was not inferior to BM-HSCT in high-risk disease [19]. While BM is the preferred source for chronic phase, PB-HSCT has been described to be similar to BM for OS, relapse, and NRM in patients with advanced CML [20]. In our cohort, PB-HSCT was an independent beneficial factor for NRM, but not for OS or PFS.
Another independent and highly significant risk factor for OS and PFS in our extended study was advanced disease stage at HSCT (BC vs non-BC). It has been shown previously that BC at HSCT was associated with extremely poor prognosis [21,22]. In our cohort, BC at HSCT was an essential risk factor for OS. Efforts to downgrade the disease to non-BC should therefore be undertaken with TKI and, if unsuccessful, with intensive chemotherapy [23,24].
Higher donor age was a new important risk factor for outcome after HSCT in advanced CML and particularly important in HSCT from unrelated donors, where donors may be chosen by younger age. The results published to date are conflicting. Donor age >30 years was not prognostic for survival in patients with chronic phase CML after matched related HSCT [25]. Another analysis on patients with AML did not find donor age to be a risk factor for OS or DFS [26]. However, older donor age (≥30 years) has been associated with increased NRM in unrelated HSCT without influencing OS or LFS [27]. A further publication in standard-risk patients with heterogeneous diseases reported a beneficial effect of younger donors (<37 years) on OS and TRM [28,29]. Our study, together with a publication in myelodysplastic syndrome [30], of patients with advanced-stage CML underscores the importance of donor age ≤ 36 years for OS and DFS. This may not apply to matched related donors >60 years, where the results were comparable to younger related donors [31,32]. The increased presence of clonal hematopoiesis in healthy individuals >40 years of age [33] might influence outcome resulting in predisposition to malignant disease. NRM was favorably influenced by PB and year of HSCT. Improvement of supportive therapy and high-resolution typing after 2008 might be responsible for this effect. Improvement over time (after 2006 and after 2010) has been reported earlier [9,34].
Donor type and patient age were both important for RI. Family donors were associated with a higher incidence of relapse. Differences in minor histocompatibility antigens in unrelated HSCT leading to higher graft-versus-tumor effect may be responsible for this observation. Higher patient age may be associated with higher leukemic or non-leukemic mutations [35] and therefore might be associated with higher relapse incidence. In addition, a higher rate of reduced intensity conditioning in older patients might explain a higher RI. The high relapse incidence was caused by resistance to TKI and limited availability of further TKI generations. Optimised transcript assessment and prophylactic TKI treatment will contribute to lower RI in the future. This analysis has beside several strengths (high number of patients including elderly; high number of comorbidities, bi-center study, and long term follow-up) but also limitations considering the retrospective nature of the evaluation and the inclusion of a considerable proportion of patients prior to the TKI era. In addition, we could not evaluate our results considering the EBMT CML score, which has been shown to influence survival predominantly in BM transplanted patients and before the TKI era [36]. The Haematopoietic HCT CI score (available only after 2005) could also not be considered [37].
In conclusion, this analysis of advanced phase CML supports the use of younger donors, the highest CD34 + cell dose (highest cell dose in our cohort 19 × 10 6 /kg bw) and the need to enter a non-BC phase before HSCT. In comparison to earlier publications, results in BC have improved considerably and may be further optimized by decreasing the current RI of 45% using more frequently maintenance TKI and/or MRD-tailored DLI.