Superiority of high-dose cytarabine–based mobilisation over cyclophosphamide or plerixafor in primary CNS lymphoma

Abstract

High-dose chemotherapy with autologous stem cell transplantation (ASCT) is a key consolidative therapy for primary central nervous system lymphoma (PCNSL). However, the optimal regimen for mobilizing haematopoietic stem cells remains undefined in this population, leading to unpredictable failure rates and suboptimal cell yields. We retrospectively analysed data from 142 patients with histologically confirmed PCNSL who underwent stem-cell mobilisation before ASCT. Three mobilisation strategies were compared: high-dose cytarabine (HD-Ara-C)–based mobilisation (n = 64), plerixafor plus G-CSF (n = 60), and cyclophosphamide (CTX, n = 18). Mobilisation success, CD34⁺ cell yield, toxicity, and predictors of collection outcomes were evaluated. Patients in the HD-Ara-C–based mobilisation group achieved a 100% success rate, significantly outperforming CTX (77.8%) and steady-state mobilisation (98.3%) while delivering a markedly superior median CD34⁺ cell yield (20 × 10⁶/kg vs. 12.7 and 10 × 10⁶/kg, p < 0.001). Notably, 79.7% of patients in the HD-Ara-C–based mobilisation group were classified as “very good” mobilisers,with all “poor” mobilisers confined to the other groups. Crucially, we identified and validated the first-day peripheral blood CD34⁺ cell count as a powerful predictor of collection outcome. A threshold of > 31 cells/µL predicted successful collection (> 2 × 10⁶/kg) with an AUC value of 0.94 (98.6% specificity, 78.6% sensitivity). Higher thresholds predicted optimal and high-yield collections. Although grade 4 thrombocytopenia was more common with HD-Ara-C, it was manageable with supportive care. HD-Ara-C–based mobilisation is a highly effective strategy for PCNSL, ensuring universal success and maximizing CD34⁺ cell yields. The first-day CD34⁺ cell count provides a robust, real-time tool for guiding apheresis. HD-Ara-C–based mobilisation should be considered the preferred regimen for fit patients, with steady-state mobilisation as an alternative for selected cases.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00277-026-06749-z.

Introduction

Primary central nervous system lymphoma (PCNSL) is a rare and aggressive extranodal non-Hodgkin lymphoma confined to the brain, leptomeninges, spinal cord, or eyes. Its incidence is increasing among immunocompetent individuals older than 60 years, accounting for approximately 2–3% of all primary brain tumours [1, 2]. Despite high-dose methotrexate (HD-MTX)-based first-line regimens (often with cytarabine/rituximab), 20–25% of patients experience primary refractory disease, and nearly 50% of patients relapse [3, 4]. Compared with whole-brain radiotherapy, autologous stem cell transplantation (ASCT) consolidation significantly improves remission depth and long-term survival, particularly in fit patients, while reducing neurocognitive toxicity [2, 5, 6]. Recent prospective clinical trials reported impressive efficacy for ASCT consolidation, with overall response rates (ORRs) ranging from 69% to 87% and 2-year overall survival (OS) rates ranging from 70% to 90% [7–9].

Successful ASCT critically depends on efficient peripheral blood stem cell (PBSC) mobilisation and collection. The collection of sufficient high-quality autologous PBSCs relies on the successful mobilisation of haematopoietic stem cells (HSCs) from the bone marrow niche into circulation. To ensure successful multilineage engraftment after transplantation, a minimum dose of 2 × 10⁶ CD34 + cells per kilogram (kg) body weight is needed, while the optimal dose is 5 × 10⁶ CD34⁺ cells/kg [10–12]. Mobilisation failure (CD34⁺ yield < 2 × 10⁶ cells/kg) prevents 15–30% of lymphoma patients from undergoing ASCT [13–18]. In elderly PCNSL patients, this challenge is exacerbated by age-related haematopoietic compromise and cumulative side effects from therapeutic exposure. Compared to younger systemic lymphoma populations, this highlights the importance of developing tailored mobilisation approaches to facilitate ASCT access for this clinically vulnerable group.

PCNSL mobilisation lacks standardized regimens and relies on approaches adapted from systemic lymphomas. While steady-state mobilisation with granulocyte colony-stimulating factor (G-CSF) + plerixafor offers convenience and lower myelotoxicity, it frequently results in suboptimal CD34⁺ cell yield [19–21]. Chemomobilisation using cyclophosphamide plus G-CSF (CTX/G-CSF), though effective in systemic lymphomas and multiple myeloma for its higher success rates and antitumor activity, has poorly defined efficacy in PCNSL following HD-MTX induction [17, 22, 23].

Cytarabine-based mobilisation has been increasingly incorporated into PCNSL protocols, exemplified by R-cytarabine consolidation/mobilisation after R-MPV induction [7, 24–26]. However, these studies lack critical data on CD34⁺ yields and failure rates. High-dose cytarabine (HD-Ara-C), a therapeutic cornerstone with excellent CNS penetration, induces profound reversible myelosuppression that may activate endogenous mobilisation pathways, suggesting a unique dual effect of simultaneous disease control and mobilisation facilitation [27–29]. While intermediate-dose cytarabine (ID-Ara-C)/G-CSF demonstrates established safety and superiority over CTX/G-CSF in systemic lymphomas and multiple myeloma [30–33], the specific efficacy of HD-Ara-C for stem cell mobilisation in PCNSL remains unevaluated, with robust comparative data notably absent.

Consequently, a comprehensive comparison of the three predominant mobilisation strategies—steady-state, CTX/G-CSF, and HD-Ara-C–based mobilisation—in PCNSL is imperative for evaluating their relative efficacy, safety, and applicability. We designed this study to generate comparative evidence for these approaches, with a focus on the dual-effect potential of HD-Ara-C-based mobilisation. Our objectives were to provide data-driven guidance for optimizing HSC harvest efficiency, reducing mobilisation failure, expanding access to consolidative ASCT, and ultimately improving long-term survival in patients with this aggressive malignancy.

Methods

Study design and patient population

We conducted a retrospective cohort analysis of 148 consecutive patients with histologically confirmed PCNSL who underwent autologous haematopoietic stem cell mobilisation at the Department of Hematology, Huashan Hospital affiliated to Fudan University between September 1, 2019, and May 31, 2025. All patients were aged 18–70 years and planned to receive high-dose chemotherapy followed by ASCT. The exclusion criteria included concurrent systemic malignancies, severe cardiopulmonary dysfunction contraindicating transplantation procedures, or patient refusal to participate in the mobilisation protocol.

Mobilisation protocols

HD-Ara-C–based mobilisation: Patients received cytarabine 3 g/m² every 12 h for two consecutive days, followed by G-CSF support. Among patients who entered mobilisation immediately after achieving complete remission with a first-line HD-MTX–based induction regimen, consolidation HD-MTX was delivered during the same admission. In contrast, patients who attained remission only after second- or later-line treatment—many of whom were clinically MTX-refractory—did not receive HD-MTX during mobilisation. In these cases, mobilisation consisted of HD-Ara-C alone, or HD-Ara-C combined with a Bruton tyrosine kinase(BTK) inhibitor in accordance with contemporary salvage practice.

Steady-state mobilisation: Patients received subcutaneous G-CSF monotherapy (10 µg/kg/day for five consecutive days). Plerixafor was administered on Day 4 of G-CSF treatment.

CTX-based regimen: Patients received a single intravenous dose of CTX (3 g/m²) over 2 days. G-CSF (10 µg/kg/day subcutaneously) was initiated during subsequent bone marrow suppression and maintained until leukapheresis was completed.

Stem cell collection and yield assessment

Leukapheresis was initiated when the PB CD34⁺ cell count reached ≥ 20/µL. Collections were performed using a continuous-flow blood cell separator (OPTIA Spectra, Terumo BCT). Mobilisation failure was defined as either inability to commence collection (persistent PB CD34⁺ <20/µL) or failure to collect the minimum target dose of ≥ 2 × 10⁶ CD34⁺ cells per kilogram of patient body weight. The collected stem cell products were subjected to quantitative analysis via flow cytometry. The total CD34⁺ cell yield, expressed as cells × 10⁶ per kg, was categorized into four tiers: poor (≤ 2 × 10⁶/kg), suboptimal (> 2 to ≤ 5 × 10⁶/kg), good (> 5 to ≤ 10 × 10⁶/kg), and very good (> 10 × 10⁶/kg).

Predictor variables and toxicity

The PB CD34⁺ cell count (cells/µL) and the percentage of CD34⁺ cells among nucleated cells were measured before the first planned apheresis day (Day 1). During mobilisation, daily monitoring included complete blood counts (including absolute neutrophil count), kidney and liver function tests, and infection markers. Febrile neutropenia (FN) was defined as a fever ≥ 38.3 °C with an absolute neutrophil count (ANC) < 0.5 × 10⁹/L. Blood-related side effects were graded using CTCAE version 4.0.

Engraftment and survival analyses

Among 137 successfully mobilized patients, 92 underwent ASCT. Neutrophil engraftment was defined as the first of three consecutive days with an ANC ≥ 0.5 × 10⁹/L without G-CSF support. Platelet engraftment was defined as the first day on which platelet counts ≥ 20 × 10⁹/L were achieved without support from transfusions for seven consecutive days. Engraftment kinetics were assessed in this cohort.

Progression-free survival (PFS) was assessed in a subset of 65 patients who underwent ASCT before January 30, 2025, with follow-up censored at July 31, 2025. PFS was defined as the time interval from the date of ASCT to the date of disease progression, death, or last follow-up. Kaplan–Meier analysis was employed to compare PFS across different mobilisation regimens and CD34⁺ cell doses (stratified at thresholds of 5 × 10⁶/kg and 10 × 10⁶/kg).

Statistical analysis

Continuous data are presented as medians with interquartile ranges (IQRs), and between-group comparisons were performed using the Kruskal–Wallis test. Categorical variables are expressed as numbers (percentages) and were analysed using chi-square or Fisher’s exact tests. Receiver operating characteristic (ROC) curve analysis established optimal cut-off values for Day 1 peripheral blood CD34⁺ levels to predict three collection endpoints: ≥2 × 10⁶ (minimum target), ≥ 5 × 10⁶ (optimal yield), and ≥ 10 × 10⁶ (high yield) CD34⁺ cells/kg. All the statistical tests used two-sided p values, with significance defined as p < 0.05. Analyses were conducted using SPSS (v26.0; IBM Corp.), R with the ggplot2, pROC, survival, and survminer packages in RStudio (2024.12.1), and GraphPad Prism (v10.1.2).

Results

Patient characteristics and mobilisation outcomes

During the study period, 148 consecutive PCNSL patients underwent HSC mobilisation. Six patients receiving G-CSF monotherapy who did not meet the plerixafor administration criteria were excluded, resulting in the identification of 142 evaluable patients (Fig. 1). The cohort included 60 females (42.3%) and 82 males (57.7%), with a median age of 53 years (range: 19–67). Patients were stratified into three mobilisation regimens: steady-state (G-CSF + plerixafor, n = 60, 42.3%), HD-Ara-C–based (n = 64, 45.1%), and CTX-based (n = 18, 12.7%). Baseline characteristics, including sex distribution, prior chemotherapy lines, prior immunomodulatory therapy, and previous whole-brain radiotherapy, were balanced across groups (Table 1).

Fig. 1.

Study flowchart. Schematic representation of patient selection, mobilisation strategy allocation, and primary outcome assessment. Abbreviations: PCNSL, primary central nervous system lymphoma; G-CSF, granulocyte colony-stimulating factor; CTX, cyclophosphamide; HD-Ara-C, high-dose cytarabine; ROC, receiver operating characteristic; CD34⁺, CD34-positive haematopoietic stem/progenitor cells

Table 1.

Patients’ characteristics and baseline statistics

	Steady-State (n = 60)	CTX (n = 18)	HD-Ara-C (n = 64)	p
Age, median (range)	56 (20–67)	44(32–55)	52.5(19–67)	0.0004
Gender, n (%)				0.71
Male	34 (56.7)	12 (66.7)	36 (56.3)
Female	26 (43.3)	6 (33.3)	28 (43.7)
Number of previous therapy regimens, n (%)				0.43
1	47 (78.3)	16 (88.9)	47 (73.4)
≥2	13 (21.7)	2 (11.1)	17 (26.6)
Prior IMiD therapy, n (%)				0.34
Present	5 (8.3)	0	2 (3.1)
Absent	55 (91.7)	18 (100)	62 (96.9)
Prior WBRT, n (%)				0.22
Present	5 (8.3)	2 (11.1)	2 (3.1)
Absent	55 (91.7)	16 (88.9)	62 (96.9)

n Number; IMiD immune modulatory drugs; CTX cyclophosphamide; HD-Ara-C high-dose cytarabine; WBRT whole brain radiotherapy

All mobilisation regimens were administered according to the planned dose-dense schedules. In the CTX group, all patients received the full planned dose. In the HD-Ara-C group, most patients received the full cytarabine dose, with only six patients (9.4%) receiving a modest dose reduction (80% of the planned dose) due to prior chemotherapy–related marrow suppression. No major deviations from intended mobilisation protocols occurred.

Apheresis efficiency by mobilisation regimen

Significant differences in mobilisation success and apheresis duration were observed between the groups. Patients in the HD-Ara-C–based group achieved 100% successful mobilisation (64/64), with 82.8% (53/64) reaching the target CD34⁺ yield in a single apheresis session. In the G-CSF + plerixafor group, one mobilisation failure occurred (1.7%), whereas 78.3% (47/60) of the participants completed collection on a single day. CTX demonstrated the highest failure rate (22.2%, 4/18), although 72.2% (13/18) of successfully mobilized patients required just one collection. Overall mobilisation failure occurred in 3.5% (5/142) of patients, predominantly in the CTX group (4/5 patients; Table 2).

Table 2.

Haematopoietic stem cell mobilization and harvest characteristics

	Steady-State (n = 60)	CTX (n = 18)	HD-Ara-C (n = 64)	p
Number of aphereses, n (%)				0.03
Not collected	1 (1.7%)	3 (16.7%)	0
One	47 (78.3%)	13 (72.2%)	53 (82.8%)
Two	12 (20%)	2 (11.1%)	11 (17.2%)
Total CD34⁺ yield (×10⁶ cells/kg), median [IQR]	10 (6.3–14.6)	12.7 (5.0–15.6)	20 (11.4–33.5)	< 0.0001
Total collection volume (mL), median [IQR]	594 (504–684)	340 (323–394)	377 (286–450)	< 0.0001
Yield Category, n (%)				< 0.0001
Poor (≤ 2 × 10⁶ cells/kg)	1 (1.7)	4 (22.2)	0
Less-than-optimal (2–5 × 10⁶ cells/kg)	9 (15)	2 (11.1)	6 (9.4)
Good (5–10 × 10⁶ cells/kg)	22 (36.7)	2 (11.1)	7 (10.9)
Very good (≥ 10 × 10⁶ cells/kg)	28 (46.6)	10 (55.6)	51 (79.7)
Day1 CD34⁺ Yield (×10⁶ cells/kg), median [IQR]	7.5 (4.4–12.1)	12.4 (4.4–15.6)	17.4 (7.7–28.7)	< 0.0001
Day 1 PB CD34⁺/µL, median [IQR]	111.8 (82.8–111.8)	80.6 (20.1–225.2)	152.9 (72.8–357.3)	0.04
Day 1 PB CD34⁺percentages (%), median [IQR]	0.16 (0.1–0.2)	0.76 (0.14–1.5)	2.3 (1.1–6)	< 0.0001
Day 1 PB TNC, median [IQR]	71005.1 (51711.6–89391.3)	14028.7 (10062.7–22072.8)	6429.06 (2988.1–20571.3)	< 0.0001

n number; IQR interquartile range; PB peripheral blood; CTX cyclophosphamide; HD-Ara-C high-dose cytarabine; TNC Total nucleated cell

CD34⁺ cell yield by mobilisation regimen

Patients in the HD-Ara-C–based mobilisation group achieved substantially greater CD34⁺ cell harvests (P < 0.0001). The median CD34⁺ yield was 20 × 10⁶ cells/kg (IQR: 11.4–33.5) in the HD-Ara-C group versus 12.7 × 10⁶ (IQR: 5–15.6) in the CTX group and 10 × 10⁶ (IQR: 6.3–14.6) in the steady-state group (Fig. 2 A). This yield advantage was achieved despite significantly smaller collection volumes: the median volume was 377 mL (IQR: 286–450) in the HD-Ara-C group versus 594 mL (IQR: 504–684) in the steady-state group and 340 mL (IQR: 323–394) in the CTX group (P < 0.0001; Fig. 2B), indicating superior cell concentration efficiency.

Fig. 2.

Impact of the mobilisation regimen on haematopoietic stem cell yield in PCNSL patients. (A) Total CD34⁺ cell yield (×10⁶ cells/kg). (B)Total collection volume (mL). Boxes represent the median and interquartile range. (C) Proportional distribution of mobilisation efficacy categories by regimen. (D) Absolute patient counts within each mobilisation efficacy category per regimen. Category definitions: Poor:<2 × 10⁶/kg; Less-than-optimal: 2–4.9× 10⁶/kg; Good: 5–9.9× 10⁶/kg; Very good: ≥10 × 10⁶/kg. *p < 0.05, **p < 0.01, ***p < 0.001. Abbreviations: HD-Ara-C: high-dose cytarabine; CTX: cyclophosphamide

When categorized by clinical yield thresholds (Fig. 2C-D), HD-Ara-C–based mobilisation resulted in 0% poor mobilizers (≤ 2 × 10⁶ cells/kg), 9.4% (6/64) less-than-optimal (2–5 × 10⁶), 10.9% (7/64) good (5–10 × 10⁶), and 79.7% (51/64) very good mobilizers (≥ 10 × 10⁶). The steady-state treatment resulted in 1.7% (1/60) poor, 15.0% (9/60) less-than-optimal, 36.7% (22/60) good, and 46.6% (28/60) very good mobilizers. CTX treatment resulted in a paradoxical distribution: 22.2% (4/18) were poor mobilizers, yet 55.6% (10/18) were very good mobilizers. Among all poor mobilizers (n = 5), 80% (4/5) originated from the CTX group. Notably, 57.3% (51/89) of all very good mobilizers were in the HD-Ara-C group.

Predictive value of day 1 PB CD34⁺ parameters

On Day 1, 142 patients in the HD-Ara-C–based mobilisation group showed significantly higher CD34⁺ counts and percentages in PB compared to the steady-state and CTX groups. HD-Ara-C mobilizers had a median CD34⁺ count of 152.9 cells/µL and a percentage of 2.3% versus 111.8 cells/µL and 0.16% in the steady-state group and 80.6 cells/µL and 0.76% in the CTX group (p < 0.05) (Table 2; Fig. 3A-C).

Fig. 3.

Day 1 PB haematopoietic stem cell parameters by mobilisation regimen and efficacy category. (A–C) Comparison of Day 1 parameters across mobilisation regimens (n=142 patients): (A) CD34⁺cell count (cells/μL) across regimens; (B) TNC count across regimens; (C) CD34⁺ cell percentage among TNCs across regimens; (D–F) Association between daily apheresis parameters and total collected CD34⁺ cell yield (stratified by efficacy category; n=167 daily samples from 142 patients, including 25 patients with 2-day collections): (D) CD34⁺ cell count (cells/μL) by efficacy category; (E) TNC count by efficacy category; (F) CD34⁺ cell percentage among TNCs by efficacy category; efficacy categories: poor (<2 × 10⁶ CD34⁺ cells/kg), less-than-optimal (2–4.9.9 × 10⁶/kg), good (5–9.9.9 × 10⁶/kg), and very good (≥10 × 10⁶/kg). *p ˂ 0.05, **p ˂ 0.01, ***p ˂ 0.001, ns, not significant. Abbreviations: HD-Ara-C: high-dose cytarabine; CTX: cyclophosphamide; PB: peripheral blood; TNC: total nucleated cell

After the 167 apheresis procedures were analysed, the Day 1 PB CD34⁺ parameters strongly correlated with the final yield categories, increasing progressively from poor (25.35 cells/µL; 0.07%) to very good mobilizers (186.58 cells/µL; 1.44%) (Table 3; Fig. 3D-F).

Table 3.

Day 1 PB Haematopoietic stem cell parameters by efficacy category

Yield Category	Poor (≤ 2 × 10⁶ cells/kg)	Less-than-optimal (2–5 × 10⁶ cells/kg)	Good (5–10 × 10⁶ cells/kg)	Very good (≥ 10 × 10⁶ cells/kg)	p
PB CD34/µL median [IQR]	25.35 (12.3–61.39)	58.47 (41.71–81.61)	107.31 (76.41–107.31)	186.58 (120.24–366.77)	< 0.0001
PB CD34⁺percentages (%), median [IQR]	0.07 (0.04–0.18)	0.18 (0.11–0.56)	0.2 (0.12–0.99)	1.44 (0.44–3.76)	< 0.0001
PB TNC, median [IQR]	64304.92 (11716.23–86677.65)	30169.22 (11557.47–59736.9)	48880.86 (11534.78–71033.29)	18857.45 (6110.46–61904.02)	0.109

IQR Interquartile Range; PB peripheral blood; TNC Total nucleated cell

ROC analysis confirmed that the Day 1 PB CD34⁺ parameters accurately predicted different yield thresholds. For a yield of ≥ 2 × 10⁶ CD34⁺ cells/kg, a PB CD34⁺ count > 31 cells/µL achieved an AUC value of 0.94 (sensitivity 0.986; specificity 0.786). Combining a count > 31 cells/µL with a percentage > 0.18% maintained an AUC value of 0.956 (95% CI: 0.908–1). For ≥ 5 × 10⁶ CD34⁺ cells/kg, a count > 86.2 cells/µL achieved an AUC value of 0.903 (sensitivity 0.842, specificity 0.86), with a combination threshold (count > 86 cells/µL + percentage > 0.19%), yielding an AUC value of 0.9 (95% CI: 0.84–0.959). For ≥ 10 × 10⁶ CD34⁺ cells/kg, a count > 126.7 cells/µL had an AUC value of 0.86 (sensitivity 0.74, specificity 0.846), with a combination threshold (count > 126.7 cells/µL + percentage > 0.269%), yielding an AUC value of 0.868 (95% CI: 0.813–0.924) (Fig. 4).

Fig. 4.

Predictive value of Day 1 PB CD34⁺ cell parameters for stem cell yield thresholds. (A) ROC analysis of the CD34⁺ cell count (cells/μL) and CD34⁺ cell percentage for predicting yield ≥2× 10⁶/kg. (B) Combined predictive model (CD34⁺ cell count + percentage) for yield ≥2× 10⁶/kg. (C) ROC analysis for yields≥5 × 10⁶/kg. (D) Combined model for yield ≥5 × 10⁶/kg. (E) ROC analysis for yields≥10 × 10⁶/kg. (F) Combined model for yield ≥10 × 10⁶/kg. Clinical thresholds: Minimum adequate (≥2 × 10⁶/kg), Optimal (≥5 × 10⁶/kg), Very good (≥10 × 10⁶/kg)

Safety profile and engraftment outcomes

Mobilisation-related toxicities included febrile neutropenia in 44.4% (8/18) of CTX patients and 35.9% (23/64) of patients in the HD-Ara-C–based mobilisation group. Grade 4 thrombocytopenia occurred in 50.0% (9/18) of CTX patients and 65.6% (42/64) of HD-Ara-C patients, although no significant bleeding events or severe infections occurred in either group (Fig. 5A-B). To further evaluate clinically significant thrombocytopenia, we assessed the proportion of patients whose platelet counts decreased to < 10 × 10⁹/L, a widely accepted transfusion threshold. This occurred in 18.7% of HD-Ara-C–based mobilisation patients and 5.6% of CTX patients (Fig. 5 C), which aligns with the observation that platelet transfusion requirements were generally low and typically limited to one unit when needed.

Fig. 5.

Treatment-emergent adverse events and haematopoietic engraftment following stem cell mobilisation and autologous transplantation in PCNSL patients. Stacked bar graphs compare regimen-related toxicity between the CTX and HD-Ara-C mobilisation regimens. (A) Incidence of febrile neutropenia (FN), defined as an ANC<0.5 × 10⁹/L with fever ≥38.3 °C. (B) Incidence of grade 4 thrombocytopenia, defined as a platelet count <25 × 10⁹/L. (C) Proportion of patients whose platelet count decreased to <10 × 10⁹/L. (D) Neutrophil engraftment, defined as an ANC≥0.5 × 10⁹/L for 3 consecutive days; (E) Platelet engraftment, defined as a PLT count ≥20 × 10⁹/L for 7 consecutive days without transfusion support. Median engraftment times with interquartile ranges are shown.Abbreviations: HD-Ara-C, high-dose cytarabine; CTX, cyclophosphamide; FN, febrile neutropenia; ANC, absolute neutrophil count; PLT, platelet; ns, not significant.

Among the 92 patients who underwent ACST, the engraftment kinetics were comparable across the mobilisation regimens. The median time to neutrophil engraftment was 10 days in all three groups. Similarly, platelet engraftment did not significantly differ among the following groups: the steady-state group at 10 days, the HD-Ara-C group at 9 days, and the CTX group at 8 days (Fig. 5D-E).

Among 65 PCNSL patients who underwent ASCT before January 30, 2025, PFS was not significantly associated with the mobilization regimen or the dose of infused CD34⁺ cells at thresholds of 5 × 10⁶ cells/kg (p = 0.68) or 10 × 10⁶ cells/kg (p = 0.59) (Figure S1). The interpretation of CTX outcomes was limited by its small cohort size (n = 2). Notably, while HD-Ara-C–based mobilisation achieved superior CD34⁺ yields, it conferred no PFS advantage over steady-state mobilisation, suggesting that increased stem cell harvest does not correlate with improved survival in this cohort.

Discussion

This comparative study fundamentally repositioned HD-Ara-C–based mobilisation as the cornerstone for HSC mobilisation in PCNSL patients undergoing ASCT. While large-scale mobilisation studies in PCNSL are limited, and prior work using HD-Ara-C (at 3 g/m² IV q12h × 4 doses) noted manageable toxicities but lacked detailed yield data [34], our findings demonstrate that HD-Ara-C, leveraging its dual role as a core induction therapy and mobilizing agent, achieved complete elimination of mobilisation failure alongside substantially increased CD34⁺ cell yields. The observed 100% success rate markedly contrasts with the significant failure rates seen with traditional approaches such as G-CSF plus plerixafor or CTX. This pronounced difference in efficacy highlights key biological insights specific to PCNSL.

The extranodal nature and typically preserved bone marrow reserve in PCNSL appear to create a unique microenvironment in which the profound myelosuppression and subsequent rebound induced by HD-Ara-C–based induction therapy are exceptionally effective for HSC mobilisation, surpassing even cytokine-driven strategies incorporating plerixafor. Reflecting the emerging recognition of this dual benefit during the study period, our centre progressively replaced CTX with HD-Ara-C as the preferred chemotherapy-based mobilisation strategy after 2019–2020. The high success rate and CD34⁺ cell yield observed in our comparative analysis strongly support the potential benefits of HD-Ara-C and suggest that it is particularly well suited for this patient population. These findings provide new directions for optimizing HSC mobilisation strategies in PCNSL.

The haematologic toxicity profile associated with HD-Ara-C, notably the high rate of grade 4 thrombocytopenia, requires careful interpretation. This myelosuppression was not an added burden of a separate mobilisation regimen but an inherent consequence of the high-dose therapy essential for effective PCNSL treatment; mobilisation was strategically integrated into this necessary therapeutic phase. Importantly, although thrombocytopenia was frequent, most patients required only a single unit of platelet transfusion, and no major bleeding events or severe infections occurred, indicating that these adverse effects were clinically manageable with standard supportive care, including transfusions and prophylaxis with antimicrobial agents. This manageable toxicity profile must be weighed against the significant clinical drawbacks of mobilisation failure observed with other strategies. The failure rates of 1.7% with steady-state mobilisation and 22.2% with CTX mobilisation frequently necessitated salvage chemotherapy attempts, which introduced delayed toxicity and treatment effects, potentially increasing the overall burden on patients.

A critical secondary finding was the validation of the absolute PB CD34⁺ count on the first apheresis day (Day 1) as a robust, clinically practical predictor of final yield across all strategies. We established highly accurate PCNSL-specific thresholds: >31 cells/µL predicted achieving > 2 × 10⁶ CD34⁺ cells/kg (minimum adequate dose) with outstanding specificity (98.6%) and sensitivity (78.6%) (AUC 0.94). This threshold’s high specificity is particularly valuable, allowing confident cancellation of futile apheresis (≤ 31 cells/µL) and prompt initiation of rescue. Higher thresholds predicted better yields: >86 cells/µL for optimal (> 5 × 10⁶/kg) and > 127 cells/µL for very good (> 10 × 10⁶/kg) collections. The CD34⁺ cell percentage was consistently inferior to the absolute count, and combining the metrics offered no substantial improvement. Thus, the absolute PB CD34⁺ count is the most reliable biomarker for real-time apheresis decisions. Attaining higher CD34⁺ cell doses extended beyond collection, conferring significant clinical benefits. Substantial evidence confirms that infused doses ≥ 5 × 10⁶/kg accelerate neutrophil/platelet engraftment and reduce transplant-related complications while improving survival outcomes [16, 35–37]. In this study, HD-Ara-C resulted in 79.7% of patients achieving yields ≥ 10 × 10⁶/kg. Such robust harvests provide operational advantages, including a strategic reserve for potential secondary transplantation or salvage therapy should clinical needs arise.

For PCNSL patients in complete remission with adequate financial resources, steady-state mobilisation using G-CSF combined with plerixafor represents a clinically appropriate alternative. This regimen has the advantage of a predictable safety profile and low mobilisation failure rate (1.7%), positioning it as a preferential replacement for CTX in HD-Ara-C-ineligible patients. Compared with the substantially higher failure risk of CTX (22.2%), the G-CSF/plerixafor approach avoided chemotherapy-related toxicity while ensuring reliable stem cell harvest. Importantly, for older patients or those with significant comorbidities, the intensive myelosuppressive HD-Ara-C–based regimen may increase the risk. In this population, plerixafor plus G-CSF may provide a safer and effective mobilisation alternative, allowing adequate CD34⁺ cell collection while minimizing haematologic toxicity. Individualized regimen selection on the basis of age, performance status, and tolerance is therefore recommended, balancing efficacy with safety considerations. Beyond PCNSL, the universal mobilisation success of HD-Ara-C–based mobilisation and superior CD34⁺ cell yield provide preliminary evidence supporting its broader exploration in aggressive B-cell lymphomas such as DLBCL. However, well-designed clinical trials remain essential to validate the safety and efficacy of these strategies in diverse populations before clinical implementation.

Critically, neither the mobilisation strategy nor the CD34⁺ cell dose (> 5 × 10⁶/kg or > 10 × 10⁶/kg) influenced PFS, but these findings were confounded by heterogeneity in conditioning regimens due to thiotepa shortages. Among 20 relapsed patients, 55% received suboptimal busulfan-cyclophosphamide-etoposide (BuCyE)-based conditioning instead of thiotepa-based therapy, introducing significant bias given the established efficacy of thiotepa. This likely obscured the potential effects of graft characteristics on survival. Nonetheless, HD-Ara-C retains primary clinical value by overcoming PCNSL mobilisation barriers (overall failure: 3.5%, predominantly CTX), ensuring transplant access. Our data support the use of HD-Ara-C for maximizing collection success, which is complemented by the validated PB CD34⁺ cell predictive model. Future studies that require uniform thiotepa-based conditioning are essential for definitively assessing survival impacts.

This study has several limitations. First, it is a retrospective, single-centre analysis, and the evolution of practice during the enrolment period led to a smaller CTX group. Second, survival data post-ASCT were not the primary focus, and heterogeneity in conditioning regimens may have introduced bias. Third, in a subset of patients with relapsed or refractory PCNSL, HD-Ara-C was combined with a BTK inhibitor as salvage therapy. This off-label regimen reflected contemporary local salvage practice at our centre and was supported by NCCN guideline–listed treatment options for relapsed or refractory disease [38]. Fourth, spleen size was not routinely monitored in our cohort. Recombinant human granulocyte-colony stimulating factor (rhG-CSF) is widely used to mobilize autologous peripheral blood stem cells (PBSCs) for autologous or allogeneic transplantation. Such mobilisation may lead to transient spleen enlargement [39, 40], and rare cases of spontaneous splenic rupture have been reported. However, recent studies on PBSC mobilisation in lymphoma and multiple myeloma have not routinely described spleen-related changes [16, 23, 33, 41, 42]. Although we did not perform systematic ultrasonographic monitoring, no clinically significant splenic events were observed. Future prospective studies incorporating routine imaging could help clarify the true incidence and clinical relevance of G-CSF–related splenic enlargement.

In conclusion, HD-Ara-C–based mobilisation has achieved universal success and maximized CD34⁺ cell yields in PCNSL patients, effectively integrating therapeutic cytoreduction with efficient stem cell collection. It outperformed both CTX- and plerixafor-based regimens, despite manageable haematologic toxicity. The validated Day 1 PB CD34⁺ cell thresholds (> 31/86/127 cells/µL) offer valuable real-time guidance for making informed apheresis decisions. Strategic integration of HD-Ara-C–based mobilisation and application of these predictive thresholds can significantly increase the reliability, efficiency, and overall success of stem cell collection, providing new insights for the use of ASCT for PCNSL treatment.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Q.L., Z.L.and L.Z. were responsible for data collection, statistical analysis, figure preparation, and manuscript writing. Q.L., Z.L. and B.C. designed the study and screened eligible patients. Y.M., J.M., and M.Z. contributed to patient recruitment and data extraction. All authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Funding

The study was supported by the Shanghai Shenkang Clinical Innovation Project (SHDC12020112). This study was also funded by Fudan University Huashan Hospital Basic Research Cultivation Program(2024JC057).

Data availability

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The study was conducted in accordance with the Declaration of Helsinki, local legislation, and institutional requirements, and was approved by the Ethics Committee of Huashan Hospital, Fudan University.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.