Minimal Residual Disease as a Predictive Factor for Relapse after Allogeneic Hematopoietic Stem Cell Transplant in Adult Patients with Acute Myeloid Leukemia in First and Second Complete Remission

Allogeneic hematopoietic stem cell transplantation (allo-SCT) is potentially curative for patients with high-risk leukemia, but disease recurrence remains the leading cause of treatment failure. Our objective was to determine the impact of minimal residual disease (MRD) by any technique in adult patients with acute myeloid leukemia (AML) in morphologic first and second complete remission undergoing allo-SCT. Fifty nine patients were eligible for the study of 160 patients transplanted over ten years. For the MRD assessment we used multiparametric flow cytometry, cytogenetics and fluorescent in situ hybridization; 19 patients (32.2%) were identified as MRD positive. Patients with MRD had a consistently worse outcome over those without MRD, with 3-years leukemia-free survival (LFS) of 15.8% vs. 62.4% and overall survival (OS) of 17.5% vs. 62.3%. Relapse rate was significantly higher in MRD-positive patients; 3 years relapse rate in MRD-positive patients was 57.9% vs. 15.1% in MRD-negative patients. Detection of MRD in complete remission was associated with increased overall mortality (HR = 3.3; 95% CI: 1.45–7.57; p = 0.0044) and relapse (HR = 5.26; 95% CI: 2.0–14.0; p = 0.001), even after controlling for other risk factors. Our study showed that for patients in morphologic complete remission the presence of MRD predicts for significantly increased risk of relapse and reduced LFS and OS.

of MRD in AML patients in morphologic remission undergoing allogeneic transplant have been reported, and the clinical significance of MRD detection by a variety of laboratory tools is still not clear. The objective of this study was to determine the impact of minimal residual disease, identified by any technique, in adult patients with AML in morphologic first and second complete remission undergoing allo-SCT, and to determine whether the MRD as defined by these techniques, would be predictive for adverse outcome.

MRD Assessment
Each patient evaluable for this study underwent bone marrow (BM) biopsy prior to transplant, and the BM samples were further investigated for persistence of MRD. All BM aspirates and biopsy samples were analyzed at the UCLA Department of Pathology and Laboratory Medicine. For the MRD assessment we used multiparameter flow cytometry, cytogenetics and fluorescent in situ hybridization (FISH). Cytogenetics and FISH analyses were performed according to the standard protocols. In flow cytometry, residual tumor cells were detected using immunofluorescence of surface markers. A panel of at least three antibodies selected on the basis of the immunophenotype of the original leukemia was used. MRD was identified as a cell population showing deviation from normal antigen expression patterns compared with normal or regenerating marrow. Any level of residual disease was considered MRD positive. Patients who had tissue involvement (leukemia cutis) with negative bone marrow were also considered for MRD assessment (patient #52).

Statistical Analyses
OS was calculated from the date of transplant until death from any cause, and surviving patients were censored at last follow-up. TRM was defined as death due to causes unrelated to underlying disease. LFS was calculated from the date of transplant until death or relapse, and patients who were alive and disease-free were censored at last follow up. Patient survival curves were estimated using the Kaplan Meier (KM) method and compared across groups using the log rank test. Cumulative incidence curves for relapse were estimated while adjusting for the competing risk of mortality. Cumulative incidence curves for TRM were estimated while adjusting for the competing risk of relapse or non-transplant related mortality. Incidence curves were compared across groups using the Fine and Gray test.
Hazard ratios (HRs) for mortality, relapse, LFS and TRM were computed under the Cox model. For relapse and TRM the Cox model was expanded to a competing risk Cox model to allow for competing risk. Risk factors were compared between the MRD-positive vs. MRD-negative groups to determine whether they were comparable. Comparisons were carried out using the exact chi-square test (categorical variables), t-test (continuous variables) or Wilcoxon rank sum test for trend (ordinal variables). To evaluate the relationship between MRD and all-cause mortality after controlling for potential confounders we used the multivariate Cox regression model. No multivariate analysis was carried out for relapse or TRM due to the limited number of events. Risk factors were initially screened using univariate Cox regression. The initial multivariate model included the following five factors that were found to be significant or marginally significant (i.e., p < 0.25) based on the initial screen: initial cytogenetic, previous disease/prior SCT, age at transplant, transplant type/source and conditioning intensity. To select the final model we used the backwards stepwise procedure with p < 0.25 as the retention criterion. All analyses were done with SAS version 9.2 (Copyright © 2002-2008 by SAS institute Inc., Cary, NC, USA). A p value of ≤0.05 was considered significant in all cases.

Results
One hundred and sixty (160) AML patients underwent allogeneic stem cell transplantation, but only 59 patients fulfilled criteria to be included in the study on the basis of the: age 18-65, complete morphologic remission, and an available and evaluable bone marrow biopsy done at least 60 days prior to the transplantation with no intervening treatment, or skin biopsy where applicable (patient #52). The remaining 101 patients were not eligible for the study on the basis of age outside the study criteria (18 patients), disease status (48 patients), lack of bone marrow biopsy prior to transplantation in our database or biopsy done more than two months prior to transplantation (30 patients) and five patients who received non-myeloablative therapy. Characteristics of the study group, including transplantation characteristics and conditioning regimens used in the study are shown in Table 1. There were 19 MRD-positive patients (32.2%) identified with any of the previously mentioned methods, 10 patients in CR1 and nine patients in CR2. Patients defined by FC alone, cytogenetics+FISH, cytogenetics+FC and leukemia cutis had worse outcome than those defined by cytogenetic or FISH alone. There were no survivors in groups when two modalities identified MRD, and only one patient in the FC group survived. In the FISH-positive group both patients are alive, and in the cytogenetics group one patient died and one survived. There was no statistically significant difference in the MRD technique used in predicting the adverse outcome. The main cause of death in MRD-positive patients was relapse, seen in 11 patients (57.9%), four patients died of TRM (21%) and four survived (21%). Detailed characteristics of the MRD-positive patients are shown in Table 2.  There were a total of 32 (54.2%) deaths, of which 15 were in the MRD-positive group (eight in CR1 and seven in CR2), and 17 in the MRD-negative group (seven deaths were in CR1 and 10 in CR2). A total of 17 (32.8%) relapses occurred, of which 11 were in the MRD-positive group (five in CR1 and six in CR2) and six in the MRD-negative group (two relapses in CR1 and four in CR2). There were 15 (25.4%) TRM events, of which four were in the MRD-positive group and 11 in the MRD-negative group. The main cause of death was relapse in 17 patients (53.1%), GVHD in six patients (18.7%), five patients died of infection (15.6%), two patients of hemorrhage (6.2%) and two of multiorgan failure (6.2%). There are 27 (45.8%) patients alive.

Overall Survival
Median overall survival (OS) was 21.8 months (range 1.4-125.9 months), with significantly better survival in the group without MRD (p < 0.05). Overall survival was significantly lower in MRD-positive compared to MRD-negative patients (p = 0.0032). The rate of mortality was 2.97 times greater among patients with MRD vs. those without MRD (HR = 2.97; 95% CI: 1.44-6.12). The actuarial OS at 3-years was 48.6% in the entire study group, 17.5% in MRD-positive group vs. 62.3% in MRD-negative group. For MRD-negative group, the 5 year survival was 54.5% and 10 year survival was 46.7%. Overall survival curves estimated using the Kaplan Meier (KM) method for patients in first (CR1) or second (CR2) complete remission, with (MRD+) or without (MRD−) evidence of minimal residual disease (MRD) are shown in Figure 1.

Leukemia Free Survival
Leukemia free survival (LFS) was significantly lower in MRD-positive patients compared to MRD-negative patients (p = 0.0006). The rate of relapse/mortality was 3.49 times greater among patients with MRD vs. those without MRD (HR = 3.49; 95% CI: 1.71-7.11). The estimate for LFS in 3-years was 47.2% in the entire study group, 15.8% in MRD-positive group vs. 62.4% in MRD-negative group. Leukemia-free survival curves estimated using the Kaplan Meier (KM) method for patients in CR1 or CR2, with (MRD+) or without (MRD−) evidence of minimal residual disease (MRD) are shown in Figure 2.

Relapse
The likelihood of relapse was significantly greater in patients who were MRD-positive at the time of transplant compared to MRD-negative (HR = 5.26; 95% CI: 1.98-13.97; p = 0.001). In the CR1 subset, the rate of relapse was 6.36 times greater among patients with MRD vs. those without MRD (HR = 6.36; 95% CI: 1.33-30.34; p = 0.020). In the CR2 subset, the rate of relapse

Treatment-Related Mortality
Cumulative incidence of treatment-related mortality (TRM) for patients in CR1 or CR2, with (MRD+) or without (MRD−) evidence of minimal residual disease (MRD) is shown in Figure 4. We compared the risk factors between MRD-positive and MRD-negative subsets: CR1/2, type and source of transplant, HLA matching, initial cytogenetics, previous disease/prior HSCT, conditioning protocol, conditioning intensity and age, as shown in Table 3. Most risk factors were roughly comparable in the two groups. However, groups differed by type/source of transplant (p = 0.07); 26.3% of MRD-positive patients were transplanted with unrelated CB compared to 17.5% of MRD-negative patients and were more likely to be HLA mismatched-7/8, 5/6 (21.1% vs. 7.5%). In addition, MRD-positive patients tended to be slightly older than MRD-negative patients by roughly five years (p = 0.08). It appears that MRD-positive patients tended to be more likely to have unfavorable karyotype of AML at diagnosis/initial cytogenetics (48% vs. 30%) and to have secondary AML (43% vs. 23%) compared to MRD-negative patients, although the differences did not reach statistical significance.
We compared mortality outcomes in patients with and without MRD after considering the following known and potential risk factors: CR1/2, initial cytogenetics, previous disease/prior SCT, type and source of stem cells, HLA matching, conditioning intensity and conditioning protocol. Age at transplant was also considered and analyzed as a continuous variable. Table 4 shows the univariate analysis of risk factors (HRs with the 95% confidence intervals and p-values). MRD-positive patients had a 2.97 times increased rate of mortality compared to MRD-negative patients (HR = 2.97; 95% CI: 1.44-6.12; p = 0.003). Older age was associated with a roughly 3% increase in the rate of mortality per one year increase (HR = 1.03; 95% CI: 1.00-1.07; p < 0.050). Patients transplanted with cord blood were at lower risk of adverse outcome. Unfavorable karyotype of AML at diagnosis was associated with a 1.78 times greater rate of mortality compared to intermediate/favorable cytogenetics, although this finding was not significant at p < 0.05 level.
Secondary AML diagnosis (vs. AML de novo) was associated with a 1.52 times greater rate of mortality although the difference did not reach statistical significance.
In the final multivariate model only MRD was significant at p < 0.05. None of the other factors were significant when controlled for simultaneously in the multivariate model. However, even when the factors found to be significant or marginally significant in the bivariate analysis were forced into the multivariate model, MRD-positive patients still had 3.3 times greater rate of mortality compared to MRD-negative patients. (HR = 3.3; 95% CI: 1.45-7.57; p = 0.0044).

Discussion
In this single institution study we investigated the prognostic value of MRD in adult AML patients undergoing allo-SCT in CR1 and CR2. Although this is a relatively small study, we can clearly see the prognostic impact of minimal residual disease in adult allo-SCT AML patients. Compared to patients who had no detectible residual leukemia, those who were positive for residual disease had significantly increased risk of relapse and death (HR = 3.3; 95% CI: 1.45-7.57; p = 0.0044). MRD-positive patients had considerably worse outcome compared to MRD-negative patients, with 3-years LFS of 15.8% vs. 62.4% and OS of 17.5% vs. 62.3%. Furthermore, relapse rate was significantly higher in MRD-positive patients; 3 years relapse rate in MRD-positive patients was 57.9% vs. 15.1% in MRD-negative patients. TRM incidence did not significantly vary by MRD/CR groups. Evidence of MRD correlated with other adverse risk factors in the investigated group. Older age, type/source of transplant, secondary AML and unfavorable karyotype of AML at diagnosis tended to be bivariately predictive for poor outcome. Similar to our study, Walter et al. [16] findings suggest that detection of any MRD by flow cytometry at the time of SCT defines a population of patients with AML who are at higher risk for adverse outcome with 4 times greater rate of mortality, with 2-year estimates of overall survival 30.2% and 76.6% for MRD-positive patients and MRD-negative patients, and 2-year estimates of relapse 64.9% and 17.6%, respectively.
Our study showed that for patients in morphologic CR1 and CR2, the presence of minimal residual disease at 1% or greater, as detected by sensitive flow cytometric assays, cytogenetics and FISH, predicts for a significantly increased risk of relapse and reduced OS and LFS. Patients in the CR1 subset were at higher risk of relapse/mortality due to MRD than patients in CR2. There is no doubt that a validated flow cytometry assay is an excellent tool for a rapid MRD evaluation, targeting patients across all genetic subgroups. Our study showed no statistically significant difference in techniques used to demonstrate MRD in predicting adverse outcome. Abnormal cytogenetics found at the time of morphologic CR has been shown to predict shorter overall survival and a higher relapse rate in the patients with AML in our and other studies [16,17]. The use of both conventional cytogenetics analysis and FISH, may overcome the limitation of lower type of sensitivity compared to flow cytometry, and can be an important methods for evaluation of MRD [18,19]. The previously described methods in detection of MRD in AML patients should be further investigated because each of them has its advantages and disadvantages and need further validation [6][7][8]20,21]. Some studies suggests that MRD-positive patients should be identified early after the first induction therapy and assigned alternative and salvage treatment strategies [22], although other studies prefer MRD monitoring at the end of treatment [23,24]. Perea et al. [24] showed that relapses were more common in patients with FC MRD level >0.1% at the end of chemotherapy treatment than in patients with <0.1%; cumulative incidence of relapse was 67% and 21% (p = 0.03), respectively. Different cut-off levels has been proposed by different authors, a level of >0.1% in our study and others [16,21,24,25], whereas some other studies suggests the cut-off value of 0.035% [23,26,27]. The cut-off value can give us suggestions how to proceed further with MRD-positive patients [21,28]. However, additional larger and prospective studies with all previously mentioned methods for detection of MRD, including molecular PCR methodology are required to establish a universal cut-off value to define a significant MRD level. Although PCR methodology is the most sensitive in detecting of MRD in AML, still nearly 40% of patients with AML have no cytogenetic or molecular markers suitable for PCR monitoring [9]. The use of multiple approaches simultaneously can increase the number of patients who can be studied and balance the limitations of individual methods.
Presence of MRD in pre allo-SCT AML patients defines a high risk group of patients. Further therapeutic possibilities in these patients, such as post-transplant donor lymphocyte infusions (DLI), alternative high dose conditioning regimens, adjuvant treatments, or other early therapeutic intervention should be considered. Recent reports on the success of DLI are encouraging to prevent/delay relapse in AML, achieving OS of up to 50% in selected patients [29], although other studies suggests that DLI is often associated with high rates of GVHD [30]. Second allograft is an acceptable and promising approach in the patients relapsing >6 months after the first allograft, with OS up to 40% [29,30]. Donor-derived natural killer cells can mediate beneficial graft vs. leukemia reactions, without GVHD [30]. Other agents, such as 5-azacytidine have been investigated in ongoing phase II clinical trial evaluation their efficacy to treat MRD based on a decreasing CD34-donor chimerism after allo-SCT [30].

Conclusions
Our study showed that MRD-positivity in adult AML patients in CR1 and CR2 can predict adverse outcome, with increased overall mortality and relapse, even after controlling for other risk factors. Although these findings should be confirmed in a larger study, they provide the basis for further studies that can allow standardization and simplification of the MRD techniques, which can be used to identify the patients who would need further treatment and proper therapy to be applied.