Clinical significance of day 5 peripheral blast clearance rate in the evaluation of early treatment response and prognosis of patients with acute myeloid leukemia

Background Minimal residual disease detection in the bone marrow is usually performed in patients with acute myeloid leukemia undergoing one course of induction chemotherapy. To optimize the chemotherapy strategies, more practical and sensitive markers are needed to monitor the early treatment response during induction. For instance, peripheral blood (PB) blast clearance rate may be considered as such a monitoring marker. Methods PB blasts were monitored through multiparameter flow cytometry (MFC). Absolute counts were determined before treatment (D0) and at specified time points of induction chemotherapy (D3, D5, D7, and D9). The cut-off value of D5 peripheral blast clearance rate (D5-PBCR) was defined through receiver operating characteristic (ROC) analysis. Prognostic effects were compared among different patient groups according to D5-PBCR cut-off value. Results D5-PBCR cut-off value was determined as 99.55%. Prognostic analysis showed that patients with D5-PBCR ≥99.55% more likely achieved complete remission (94.6% vs. 56.1%, P < 0.001) and maintained a relapse-free status than other patients (80.56% vs. 57.14%, P = 0.027). Survival analysis revealed that relapse-free survival (RFS) and overall survival (OS) were longer in patients with D5-PBCR ≥99.55% than in other patients (two-year OS: 71.0% vs. 38.7%, P = 0.011; two-year RFS: 69.4% vs. 30.7%, P = 0.026). In cytogenetic-molecular intermediate-risk group, a subgroup with worse outcome could be distinguished on the basis of D5-PBCR (<99.55%; OS: P = 0.033, RFS: P = 0.086). Conclusions An effective evaluation method of early treatment response was established by monitoring PB blasts through MFC. D5-PBCR cut-off value (99.55%) can be a reliable reference to predict treatment response and outcome in early stages of chemotherapy. The proposed marker may be used in induction regimen modification and help optimize cytogenetic-molecular prognostic risk stratification. Electronic supplementary material The online version of this article (doi:10.1186/s13045-015-0145-1) contains supplementary material, which is available to authorized users.


Background
Acute myeloid leukemia (AML) is a group of clinically and genetically heterogeneous diseases [1,2]. Despite treatment advancements in acute promyelocytic leukemia (M3), current treatment of AML is based on chemotherapy. Standard induction chemotherapy consists of anthracycline and cytarabine (3 + 7 regimen) can achieve the complete remission (CR) rate of approximately 75%, but outcome is uncertain because of the variability of individual genetic profile and drug sensitivity [3,4]. Intense chemotherapy or allogeneic hematopoietic cell transplantation (allo-HSCT) can benefit patients who are refractory or tend to relapse [5]. Early and easy monitoring of minimal residual disease (MRD) reflects treatment response in time and becomes an essential reference for patients with AML to optimize chemotherapy.
Multiparameter flow cytometry (MFC) has been used as a standard technique to track MRD in leukemia patients in the past decades [6]. In patients with acute lymphoblastic leukemia (ALL), the threshold of 0.01% of the bone marrow (BM) MRD is considered as the boundary of relapse predict index [7,8]. In AML, ambiguous threshold is approximately 0.1% and accuracy is approximately 10 times lower than that in ALL [9]. Moreover, BM MRD status in AML cannot be considered as an independent prognostic predictor, even though this status is considered as such in ALL [8].
Either in ALL or AML patients, the time point of BM MRD measurement is usually after CR, which may be late to determine early treatment response. Patients also hesitate to undergo frequent BM aspirations. Thus, peripheral blood (PB) blast clearance in early stage of chemotherapy has been extensively investigated. In ALL, encouraging results have been reported; for instance, the complete clearance of PB blasts within the first week of treatment may be related to CR achievement [10,11]. In patients with AML, the PB blast clearance rate (PBCR) is closely correlated with treatment response and survival, but sampling time point and cut-off value vary [12][13][14][15][16].
In this study, we assessed the prognostic value of PBCR during induction in a cohort of 96 newly diagnosed AML patients. An earlier, easier, and more accurate technique than current systems has been established to distinguish high-risk patients and to enable a prompt improvement of induction chemotherapy.
Leukemia-associated aberrant immunophenotypes (LAIPs) were identified in 72 (75%) patients with four main types: 63 cases with cross-lineage antigen expression, 6 cases with asynchronous antigen expression, 13 cases with antigen dim/strong expression, and 8 cases with antigen expression missing. The details of LAIP distribution were listed in Additional file 1: Table S1.

Determination of D5-PBCR cut-off value
The PBRR of D 5 showed the highest significant difference between early CR and NCR groups. The difference in PBRRs related to the relapse status also showed that D 5 was the first time point to appear with a definite prognostic value. Receiver operating characteristic (ROC) analysis was performed to evaluate the predictive power of PBRR on patients' complete remission. The area under the curve (AUC) of D 5 was 0.746, which was larger than that of D 7 (0.736) and D 9 (0.719),and hence showed more statistical significance. Thus, D 5 was chosen as the time point to determine D5 peripheral blast clearance rate (PBCR) cut-off value.
The day 5 PB blast reduction ratio (D5-PBRR) of 2.35 was selected as the optimal cut-off, according to the maximum sum of the sensitivity and specificity with 52.2% and 92.6% on the ROC curve, respectively ( Figure 1). A logarithmic value of 2.35 of PBRR is equal to the clearance rate of 99.55% initial peripheral blasts; thus, D5-PBCR was determined as 99.55%.
The distribution of the cytogenetic-molecular risk classification showed that more unfavorable cases were classified into the lower D5-PBCR (<99.55%) group  (Figure 2A and B), even close to that of the unfavorable-risk group (Additional file 3: Figures S1A and S1B). In the favorable-risk group, no differences were observed in distribution or survival analysis.

Discussion
Based on clinical characteristics, cytogenetic and molecular markers, more precise prognostic stratification has been established in AML and diagnosis and treatment individualization has become feasible [3]. The establishment of an effective monitoring method is essential for the evaluation of the early response and the adjustment of treatment regimens and the improvement of the prognosis as well.
MRD detection by flow cytometry has been applied to identify subclinical levels of leukemia cells and evaluate  treatment more precisely than conventional morphology; as such, this technique has been considered ideal for chemosensitivity assessment [6,[17][18][19][20][21]. In ALL, MRD level (<0.01%) and cytogenetic-molecular markers are both considered as independent outcome predictors [8].
MRD-based clinical approaches in children and adult ALL have yielded excellent results, which confirmed that MRD can be effective in risk stratification and treatment intervention [22][23][24][25].
In AML, MRD monitoring also plays an important role in the evaluation of treatment effect. Retrospective studies have demonstrated a high prognostic value of post-induction MRD level in AML. Terwijn et al. [26] and Freeman et al. [27] defined 0.1% as the MRD cut-off value; in both studies, MRD is correlated with RFS but not with OS. However, Inaba et al. [28] reported that MRD has a limited value in childhood AML if measured by MFC. Thus far, the threshold of MRD related to prognosis of AML remains controversial. MRD cannot be considered as an independent prognostic predictor in AML, in contrast to ALL [26,27,29].
Either in ALL or AML, the time points of BM-MRD monitoring are not earlier than 2 weeks after the induction begins. Although prognostic correlation is excellent, the time at which correlation is determined may be too late for early intervention of induction regimen. Thus,  In patients with ALL, practical values of PB blast assessment have been reported. Gajjar et al. [10] found that the persistence of circulating blasts after 1 week of therapy is significantly related to the worse event-free survival in childhood ALL. Atsushi et al. [11] also presented the same conclusion in patients with ALL treated with prednisolone monotherapy. Studies related to AML have also shown that a rapid decrease in peripheral leukemic burden determined through either morphology or flow cytometry [12,15] is correlated with CR and longterm survival [14,16].
Our D5-PBCR cut-off value might help optimize current cytogenetic-molecular prognostic risk stratification. In our cohort, 84.21% (16/19) of unfavorable-risk patients were classified into low D5-PBCR group (<99.55%). In the intermediate-risk groups, D5-PBCR could further distinguish the subgroup of patients with relatively poorer prognosis, the two-year estimated OS and RFS rate were significantly worse in patients with D5-PBCR <99.55% (P = 0.033, P = 0.086), which were close to that of the unfavorable-risk group. This might facilitate further treatment regimen adjustments. The intermediate-risk patients with low D5-PBCR may be recommended for strengthening induction and consolidation therapy or receiving allo-HSCT. D 5 of induction is a valuable time point of early treatment response monitoring. This time point is appropriate to strengthen induction therapy. Augmented induction using dose-escalated regimens or three-drug combination has benefited patients [30][31][32]. Holowiecki et al. [31] also suggested that the addition of cladribine to the standard induction regimen can improve the outcome of patients with AML, particularly in the unfavorable-risk group. Thus, early risk evaluation by D5-PBCR provides the basis of individualized induction therapy.
The applicability of MRD by MFC is approximately 60% to 88% in patients with AML [6,17]. In patients with no LAIPs, BM MRD evaluation is unlikely performed through MFC assessment. In this case, D5-PBCR is considered as a new effective evaluation method. Almost all patients with AML suffer from circulating blasts at diagnosis, which enables PB blast monitoring a comparatively universal way. In our study, LAIP coverage was approximately 75% (72/96). Of the 24 patients with no LAIPs, 23 showed available data of D5-PBCR. The CR rate was 100% (7/7) in the high D5-PBCR group compared with 50% (8/16) in the low D5-PBCR group (P = 0.052); this result suggested that high D5-PBCR may help predict the CR status of patients without LAIPs.
However, for the few AML patients with very low percentage or even absent of PB blasts at diagnosis, BM MRD detection plays a more important role.

Conclusions
An effective evaluation method of early treatment response was established by monitoring PB blasts through MFC. D5-PBCR cut-off value might help distinguish high-risk patients in the first week of induction; thus, prognostic predictive ability of current risk stratification can be improved and induction regimen modification can be performed.
To further establish practical and precise clinical guidance, more patients need to be accumulated and multicenter confirmation is required.

Patients and treatment protocols
A total of 96 patients with de novo AML (non-M3) were enrolled in this study from June 2011 to August 2014 in the Shanghai Institute of Hematology. The diagnosis and classification of the AML subtypes were established according to FAB [33] and WHO 2008 criteria [34]. The Ethics Committee of Ruijin hospital approved this study. All patients provided informed consent according to the Declaration of Helsinki.

Cytogenetic and molecular analysis
Cytogenetic data was available in 88 of 96 cases. The chromosomes were R-banded and/or G-banded in unstimulated BM cells after 24 h of culture. The karyotype was analyzed according to the International System for Human Cytogenetic Nomenclature (2009) [37].
More than 2.5 × 10 5 events from PB samples and 1 × 10 6 events from BM samples were required for flow cytometry detection to ensure sensitivity and accuracy. For BM samples, a cut-off value of 0.1% was determined as MRD positive.