The factors influencing clinical outcomes after leukapheresis in acute leukaemia

Leukapheresis is used for the mechanical removal of leukaemic cells in hyperleukocytosis. However, the effectiveness of leukapheresis remains unclear due to selection and confounding factors in the cohorts. We compared the effectiveness of leukapheresis among the subgroups according to either the 2016 World Health Organization classification or the number of cytogenetic abnormalities with a retrospective, single-centre study from January 2009 to December 2018. Acute myeloid leukaemia (AML, n = 212) and acute lymphoblastic leukaemia (ALL, n = 97) were included. The 30-day survival rates (95% confidence interval, 95% CI) for AML and ALL were 86.3% (81.6–90.9%) and 94.8% (90.3–99.2%), respectively. For AML, ‘primary AML with myelodysplasia-related changes’ and ‘AML with biallelic mutation of CEBPA’ showed better 30-day survival outcomes (P = 0.026) than the other subgroups. A higher platelet count after leukapheresis was associated with better 30-day survival in AML patients (P = 0.029). A decrease in blast percentage count after leukapheresis was associated with better 30-day survival in ALL patients (P = 0.034). Our study suggested that prophylactic platelet transfusion to raise the platelet count to 50 × 109/L or greater might improve clinical outcome in AML patients undergoing leukapheresis.


Patients.
Most patients with an initial WBC count equal to or higher than 100 × 10 9 /L were treated with repeated leukapheresis until the WBC count decreased below 100 × 10 9 /L during admission. Those patients received induction CTx as soon as the initial leukapheresis was completed, and a bone marrow study revealed haematologic malignancy. In this study, to avoid bias from cumulative leukapheresis procedures or concomitant CTx regimens, we collected only initial leukapheresis data from each patient.
Among 521 newly diagnosed acute leukaemia patients who required emergency leukapheresis at our institution from January 2009 to December 2018, we enrolled 309 patients who were diagnosed with AML or ALL. We excluded 128 patients who were diagnosed with another haematologic malignancy as well as 84 patients with previous diagnosis and treatment of leukaemia to avoid bias that can arise from previous medical interventions. Finally, we collected data on 212 AML patients and 97 ALL patients (Fig. 1).
Patients diagnosed before 2016 were reclassified according to the 2016 WHO classification. For reclassification, cytogenetic and molecular findings, including NPM1 and CEBPA genes, were available in our medical records.
Additionally, for AML, the ' AML not otherwise specified (AML, NOS)' group was subdivided into two subgroups according to the ordinary French-American-British (FAB) classification, namely, AML M4 and M5, which are reported to be closely related to leukostatic symptoms derived from HL 4,[15][16][17][18][19] . ' AML with myelodysplasia-related changes (AML with MRC)' is a heterogeneous group that consists of multiple subgroups confirmed by different findings 20 . Therefore, this group was subdivided into two subgroups: 'primary AML with MRC' , which exhibited de novo multilineage dysplasia or cytogenetic abnormalities, and 'secondary AML with MRC' , which is derived from a history of myelodysplastic syndrome (MDS) or myelodysplastic/myeloproliferative neoplasm (MDS/MPN).
Irrespective of this subgrouping, all patients were categorized according to the number of cytogenetic abnormalities as having a normal karyotype or having one, two, or three or more cytogenetic abnormalities (a complex karyotype). We used these two approaches in the same cohort to compare the effectiveness of leukapheresis between subgroups.
Leukapheresis. All leukapheresis procedures were initiated before administration of a cytoreductive regimen. As this study was performed at a single centre, the indications for leukapheresis and the procedure itself were consistent. A central or peripheral venous catheter was placed prior to leukapheresis. The procedure was performed at the bedside using the COBE SPECTRA (Terumo BCT, Lakewood, CO, USA) continuous-flow blood cell separator. Acid citrate dextrose was added for anticoagulation with a citrate-to-blood ratio of 1:13. To avoid a hypocalcaemic effect, 1200 mg of 3% calcium gluconate added to 1 L of Hartmann's solution was administered throughout the procedure.
All the procedures were continued until approximately 7 L of blood had been processed. Each procedure lasted for a median (range) of 170 min (100-231 min), and the median volume (range) of product collected was approximately 720 mL (300-887 mL), with a coefficient of variation less than 10.0% (Supplementary Figure S1).
Blood cell counts were performed before and after the procedure. Red blood cell (RBC) transfusion should ideally be avoided before leukapheresis 2,3,12,21 . However, to manage possible haemorrhage or bleeding, some cases of anaemia were corrected via packed RBC transfusion before the procedure to reach a haemoglobin (Hb) concentration over 7.0 g/dL 22 . In thrombocytopenia, single-donor platelets (SDPs) were transfused to reach a platelet count over 50 × 10 9 /L before the procedure. Fresh frozen plasma (FFP) was transfused whenever coagulopathy was suspected before the procedure.
'WBC count gap', 'Blast absolute count gap' and 'Blast percentage gap'. To evaluate the performance of leukapheresis, three parameters before and after the procedure were mainly used. First, 'WBC count gap (× 10 9 /L)' indicates the value obtained after subtracting the postleukapheresis WBC count from the preleukapheresis WBC count. Second, the blast absolute count was derived by multiplying the WBC count by the blast percentage. In addition, 'Blast absolute count gap (× 10 9 /L)' indicates the value obtaining by subtracting the Remission. After the initial leukapheresis procedure, most patients, excluding those who died before intensive care, received induction CTx according to the disease entity confirmed by the bone marrow study. To analyse the association between initial leukapheresis and haematologic complete remission (CR), a follow-up bone marrow study was performed within 2 months for each CTx-treated patient.

30-day survival outcome.
Most patients were hospitalized for medical treatment until discharge. The effectiveness of leukapheresis was expected to influence early mortality. Data were collected at 30 days from the date of the initial leukapheresis procedure.
Statistical methods. Categorical variables are presented as frequencies and were compared using the chi square test among the groups. If more than 20% of cells had less than the expected frequency of 5, Fisher's exact test was alternatively used for analysis. Continuous variables are presented as the median (range) and were compared using Student's t-test. Comparisons of laboratory data before and after leukapheresis in each group were performed using paired t-tests. Prognostic factors were predicted by univariate and multivariate analysis via the Cox regression method, and www.nature.com/scientificreports/ all the factors were entered into the model by the backward method. To study the factors that resulted in complete remission, univariate and multivariate analyses using logistic regression were performed for patients who received intensive CTx. In addition, 30-day survival outcomes were analysed using the Kaplan-Meier method and the log-rank test. The level of statistical significance was set at a p-value of 0.05. Data were analysed using the statistical program SPSS 24 (SPSS Inc., Chicago, IL, USA).

Results
Biological and clinical characteristics. Baseline characteristics, concomitant leukostatic symptoms, transfusion records before leukapheresis, and modality setup data for the two groups are described in Table 1.
In 212 AML patients, the median WBC count and blast percentage in the peripheral blood were 164.0 (× 10 9 /L) and 90 (%), respectively. The number of patients without cytogenetic abnormalities was greater than that of patients with cytogenetic abnormalities. The number of patients who featured at least one leukostatic symptom was 103, but the presence of any symptoms was not significantly related to CBC parameters (Supplementary Table S2). In addition, the presence of each cluster of designation (CD) marker showed no significant association with the WBC count or blast percentage (Supplementary Table S3).
In 97 ALL patients, the median WBC count and blast percentage in the peripheral blood were 204.8 (× 10 9 /L) and 93 (%), respectively. The patients with cytogenetic abnormalities outnumbered the patients without cytogenetic abnormalities. The patients who featured at least one leukostatic symptom were 62 patients; in particular, the presence of headache was significantly associated with a higher WBC count than the absence of headache, with a median of 218.3 (116.1-584.8) (× 10 9 /L) versus 194.0 (83.8-615.2) (× 10 9 /L) (P = 0.029) (Supplementary Table 1. Baseline characteristics of acute myeloid leukaemia (AML), acute lymphoblastic leukaemia (ALL) groups. WBC white blood cell, RBC red blood cell, Hb haemoglobin, PRC packed red blood cells, SDP single donor platelets, PC platelet concentrates, FFP fresh frozen plasma, ACD-A acid citrate dextrose-A, NS not significant. a Missing data in the AML (5 cases), and the ALL (1 case) groups. 'WBC count gap' and 'blast percentage gap' in AML and ALL. After initial leukapheresis, the peripheral WBC count significantly decreased in the AML and ALL groups (P < 0.001, both) (Fig. 2). The peripheral blast percentages also significantly decreased in the AML and ALL groups (P < 0.001, both).  www.nature.com/scientificreports/ Using the 2016 WHO classification, with the exception of a few subgroups that were too small to be analysed, all disease entities showed a significant reduction in WBC counts after initial leukapheresis (Supplementary  Table S4). All subgroups showed a decrease in blast percentage, with or without statistical significance.
'WBC count gap' and 'blast percentage gap' according to each CD marker. In AML, the patients whose blasts were positive for CD14 showed a significantly smaller WBC count gap between procedures than those whose blasts were negative for CD14, as follows: 72.6 (18.3-169.6) (× 10 9 /L) versus 86.0 (11.0-302.2) (× 10 9 /L) (P = 0.003) (Supplementary Table S3). This implies that the WBC count of CD14-positive patients seldom decreased compared with that of CD14-negative patients. The patients whose blasts were positive for CD13 or HLA-DR showed significantly more blast gaps, but their differences were small. Other CD markers, including CD2, CD7, CD11c, CD33, CD34, CD56, CD64, CD117 and even cytoplasmic MPO, did not show significant associations with the WBC count gap or blast percentage gap between procedures.
In ALL, no significant association between each CD marker and the above parameters was observed (Supplementary Table S3).
CR rate. Subgroup analysis for CTx recipients in the AML or ALL group was performed to predict covariates related to CR. Among 161 CTx patients with AML, 107 had CR, and 54 had non-CR.
The median WBC count gaps for the CR group and non-CR group were 75.2 and 88.2 (× 10 9 /L), respectively (Fig. 3). The median initial blast absolute counts for the CR group and non-CR group were 130.9 and 166.8 (× 10 9 /L), respectively. After leukapheresis, these parameters changed to 63.3 and 65.8 (× 10 9 /L), suggesting that the blast absolute count gap is proportional to the initial blast absolute count. In CR prediction, the smaller the blast absolute count gap was, the higher the probability of CR in our model ( Table 2).
The median initial platelet count was lower in the CR group than in the non-CR group: 45.0 and 55.0 (× 10 9 /L), respectively (P = 0.034) (Fig. 3). However, the platelet count became higher in the CR group than in the non-CR group after leukapheresis, mainly due to the effect of platelet transfusion before the procedure. In CR prediction, the smaller the platelet count gap was, the higher the probability of CR (Table 2), which was expected in our model.
In ALL, among 86 CTx patients, 56 had CR, and 30 had non-CR. There were none of the covariates that predicted CR in ALL.
In addition, there were no significant differences based on the number of cytogenetic abnormalities. When we subdivided these chromosome-based data into AML and ALL (Supplementary Tables S7 and S8), there was still no definite association between the CR rate and the number of cytogenetic abnormalities (P = 0.470 for the AML group, P = 0.393 for the ALL group).
There were no significant differences among the subgroups categorized according to the number of cytogenetic abnormalities. When these subgroups were analysed in the category of either AML or ALL ( Supplementary  Fig. S11), the result was not significant (P = 0.795, 0.389, respectively).
We analysed early mortality and categorized subgroups according to several CBC parameters. The WBC count gap, blast percentage gap, and platelet count around the time of the leukapheresis procedure were studied in the AML and ALL groups. Specific cut-off levels for each parameter were generated by receiver operating characteristic (ROC) curves.
In AML, there was no significant difference when groups were subdivided by WBC count gap with a cut-off value of 80 × 10 9 /L (P = 0.834) or blast percentage gap with a cut-off value of 1% (P = 0.890) (Fig. 4a, c). However, there were significant findings when the groups were subdivided by platelet count with a cut-off value of www.nature.com/scientificreports/ 50 × 10 9 /L (P = 0.029) (Fig. 4e). A higher platelet count around the time of leukapheresis was correlated with a higher 30-day survival rate of 89.9% (95% CI, 83.5-94.3%), compared to 79.7% (95% CI, 75.2-84.1%) for patients with a lower platelet count. In ALL, no significant findings were observed when the groups were subdivided by the WBC count gap with a cut-off value of 80 × 10 9 /L (P = 0.591) (Fig. 4b). When subdivided the groups were subdivided by blast percentage gap with a cut-off value of 1%, a higher blast percentage gap was correlated with a favourable prognosis, with a 30-day survival rate of 98.4% (95% CI, 95.9-100%), compared to 88.6% (95% CI, 82.2-94.9%) for patients with a lower blast percentage gap (P = 0.034) (Fig. 4d). There were no significant differences according to platelet count (P = 0.362) (Fig. 4f).   Figure 3. Changes in CBC parameters before and after leukapheresis in the non-CR and CR groups in the acute myeloid leukaemia (AML) group. Each error bar represents the range from the 25th to 75th percentile of each group. CBC complete blood cell, CR complete remission, WBC white blood cell.  (Table 3). In AML, the groups that were not treated with CTx or exhibited an absence of CD117 showed unfavourable 30-day survival, with hazard ratios of 8.47 (3.584-18.076) and 5.68 (2.638-12.230), respectively (P < 0.001, both). In ALL, a higher blast absolute gap or lower blast absolute count after leukapheresis predicted a favourable 30-day prognosis, with a hazard ratio of 0.869 (0.760-0.994), and this difference was statistically significant (P = 0.040). Leukapheresis frequency predicted unfavourable 30-day survival, with a hazard ratio of 3.994 (1.413-11.920), and this difference was statistically significant (P = 0.009).

Discussion
There have been some studies on the effectiveness of leukapheresis in leukaemia patients. Giles et al. 23 and Bug et al. 24 reported that leukapheresis reduced early mortality. In contrast, Chang et al. reported that leukapheresis had no significant influence on early mortality 25  Oberoi et al. pointed out that some previous studies were at high risk of selection and confounding bias 28 . A randomized controlled trial has not been conducted, but analytical studies with matched control groups have been used in an attempt to minimize bias. Nan et al. found that leukapheresis was associated with a significantly lower early mortality rate than a matched control group 29 . However, Stahl et al. were unable to detect an early mortality benefit from leukapheresis compared with matched controls 30 .
All studies mentioned above focused on AML. Other types of leukaemia may also feature HL, so leukapheresis should be taken into consideration. Choi et al. performed a propensity score-matched study for AML and ALL and did not observe significant differences in early mortality 31 .
Even in studies using a matched control group, the results have remained heterogeneous, making it difficult to perform routine leukapheresis. We posit that this heterogeneity is caused by multiple factors, including differences in detailed disease entities specified by the 2016 WHO classification as well as the number of cytogenetic abnormalities.
In our study, some patients did not receive chemotherapy, as they died before it could be administered. Excluding them brought about a shortage of patients in each subgroup, so we did not analyse remission rates by the 2016 WHO classification. There was no significant correlation between the remission rate and the number of cytogenetic abnormalities. Some studies have mentioned the association between the number of cytogenetic abnormalities and CR rates, which are inversely proportional to each other in AML [32][33][34][35] and not significant in ALL 36 . However, in our study, the subgroup with 3 or more cytogenetic abnormalities in AML and ALL showed CR rates indistinguishable from those of subgroups with fewer cytogenetic abnormalities after leukapheresis (Supplementary Tables S7 and S8).
We regarded the 30-day survival rate as the most reliable outcome. Considering the 2016 WHO classification, we noticed two findings in AML (Supplementary Fig. S9). First, there were no deaths in the ' AML with CBFB-MYH11′ or 'primary AML with MRC' subgroup. According to the 2017 European LeukemiaNet (ELN) risk stratification, all ' AML with CBFB-MYH11' patients in our cohort are considered favourable, whereas 'primary AML with MRC' patients have either intermediate or adverse risk 37 . Second, ' AML with biallelic mutation of CEBPA' had the highest WBC count of 241.9 × 10 9 /L (130.9-346.5) among all AML subgroups (Supplementary  Table S4). However, the 30-day survival rate was 93.3%, higher than all other subgroups except ' AML with CBFB-MYH11' and 'primary AML with MRC' . These data might imply that 'primary AML with MRC' and ' AML with biallelic mutation of CEBPA' may benefit more from leukapheresis than other subgroups.
In addition to disease entity, we attempted to correlate clinical outcomes with CBC parameters in AML and ALL. Intuitively, a wider range of WBCs and blast absolute count gaps after leukapheresis seemed to be related to higher CR rates. However, AML in our study, with a wider range of blast absolute count gaps, showed an odds ratio less than one for CR (Table 2). This could be interpreted as a negative effect of leukapheresis for establishing CR, but the authors thought that the reason originated from the difference in the initial blast absolute count rather than the blast absolute count gap. As shown in Fig. 3b, the non-CR group had a higher initial blast absolute count, so a wider blast absolute count gap was observed after leukapheresis, with a correlation coefficient of 0.739 (P < 0.001). Giles et al. reported that HL significantly decreased the CR rate 23 , and Chen et al. reported Table 2. Prediction for complete remission (CR) using univariate and multivariate logistic regression in acute myeloid leukaemia (AML) group. OR odds ratio, CI confidential interval. www.nature.com/scientificreports/ that HL required more chemotherapy cycles than non-HL to achieve CR 38 . Therefore, performing leukapheresis and reducing WBCs or blasts might still be more important than a greater number of WBCs or blast reduction efficiency for CR in AML. In ALL, a higher blast absolute count gap did not predict favourable CR rates. There might have been several reasons, but the authors suggest that this result is derived from the initial blast absolute count. As shown in Fig. 2b, a much higher initial blast absolute count and a relatively wider range of blast absolute count gaps are observed in ALL than in AML. Even if blast reduction by leukapheresis would also be www.nature.com/scientificreports/ helpful for achieving CR in ALL, wider dispersion of blast cell counts in ALL might result in an insignificant P-value in regression analysis. A larger reduction in WBC count by initial leukapheresis did not ensure a better survival rate (Fig. 4a, b). However, the 30-day survival rate according to blast percentage gap was significant only in ALL (Fig. 4d). The percentage is the relative parameter, and a greater than 1% blast percentage gap indicates a decrease in blast percentage among blood cells after leukapheresis. This could be interpreted in various ways. The authors conceived that reducing haematocrit in the collection port may lead to efficient blast removal while minimizing the waste of other blood cells, as blasts are positioned in the upper layer in the buffy coat.
Meta-analysis by Oberoi et al. 28 , as well as recently published studies by Stahl et al. 30 and Bewersdorf et al. 39 , showed that there were no 30-day survival benefits after leukapheresis. The difference between previous studies and this study is that transfusion before leukapheresis was considered and platelet supplementation was a covariate that influenced clinical outcomes. RBC or Hb around the time of leukapheresis did not show significant outcome differences, but platelet count was favourably associated with CR and 30-day survival in AML (Table 2 and Fig. 4e). Previously, platelet count was reported to be decreased after leukapheresis [40][41][42] , and prophylactic platelet transfusion is recommended before central venous catheter insertion for patients with platelet counts less than 20 × 10 9 /L 43,44 . However, there are no guidelines for prophylactic platelet transfusion for leukapheresis, and there is wide variation across regional and institutional guidelines. Platelets are related with immune function by activating dendritic cells, secreting granules that activate T-cells, influencing T-/B-cell response, supporting innate and adaptive immunity, influencing leukocytes, and containing Toll-like receptor transcripts [45][46][47] . As platelets are related to coagulation and immunologic functions, replenishment of platelets by transfusion in the CR group might have been related with favourable prognosis. In our study, a platelet count higher than 50 × 10 9 /L prior to leukapheresis in AML was associated with CR (Table 2) and favourable 30-day survival. Additional data and evidence are required for administration of platelet as a therapeutic modality.
Our study has several limitations. Comparison groups and subgroups without leukapheresis were not included, as most patients with an initial WBC count equal to or higher than 100 × 10 9 /L routinely underwent leukapheresis at our institution. The post-CBC sampling time interval from leukapheresis ranged from less than 1 to 24 h, which might cause variation in the results. Nevertheless, our study offers new insight into differences between disease entities based on modern concepts. Moreover, haematologic conditions prior to leukapheresis can be the sole rationale for clinical decision making.
In summary, the authors compared the effectiveness of leukapheresis among subgroups according to either the 2016 WHO classification or the number of cytogenetic abnormalities. All subgroups showed a significant reduction in WBC count after leukapheresis. In the 2016 WHO classification, the 'primary AML with MRC' and ' AML with biallelic mutation of CEBPA' subgroups showed better early survival outcomes than other AML subgroups. The statistical model showed that the blast absolute count gap was related to CR in AML. This suggested that leukapheresis may have an indirect positive impact on leukaemia treatment. Moreover, platelet count around the procedure was favourably associated with CR and the 30-day survival rate in AML. Therefore, the tumour burden that may inhibit CR could be removed via leukapheresis with transfusion of enough platelets.