Autologous stem cell transplantation for secondary central nervous system lymphoma: a multicentre retrospective analysis

Abstract

Background

Secondary Central Nervous Lymphoma (SCNSL) is an aggressive lymphoma with poor outcomes. Consensus regarding consolidation with thiotepa-based autologous transplant (ASCT) for SCNSL is lacking.

Methods

Our retrospective Singapore study examined survival outcomes post ASCT in 27 patients (median age 57 years) with SCNSL in two tertiary cancer centres from 2013 to 2024.

Results

Five (18.5%) patients had de novo SCNSL, 14 (51.9%) patients had isolated CNS relapsed SCNSL and 8 (29.6%) patients had synchronous relapsed SCNSL. At the time of ASCT, 16 (59.3%), 7 (25.9%) and 4 (14.8%) patients were in CR, PR and SD/PD respectively. Median follow up for the 22 relapsed SCNSL patients was 63 (5–109) months with a median PFS and OS of 11.5 and 30 months, respectively, and 4-year PFS and OS rates of 35.0% and 48.0%, respectively. In our analysis, females demonstrated superior median and 4-year PFS and OS compared to males, likely related to difference in age and performance status. On multivariate analysis, CR/PR at time of ASCT trended toward an improved OS (HR 7.86, p = 0.06) although this was not statistically significant.

Conclusion

Our data suggests that in secondary CNSL, chemo-sensitivity would likely confer good outcomes with thiotepa-based consolidation ASCT. Conversely, patients responding poorly to salvage chemotherapy may consider alternatives including CAR-T cell therapy.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12885-026-16531-z.

Introduction

Secondary Central Nervous System Lymphoma (SCNSL), most commonly of diffuse large B cell subtype, is an aggressive disease with poor outcomes [1]. SCNSL has a distinct pathophysiology compared with Primary CNS lymphoma, which is confined to immune-privileged sites. It may present as de novo systemic disease with concurrent CNS involvement at diagnosis, or, following prior treatment, as an isolated CNS relapse, or synchronous systemic and CNS relapse [2].

Induction therapy typically consists of systemic chemoimmunotherapy intercalated with CNS-penetrating chemotherapy, with treatment intensity and consolidation strategies guided by patient fitness [3]. Consolidation with high dose chemotherapy incorporating thiotepa followed by autologous stem cell transplant (ASCT) offers a potentially curative option for patients with chemo-sensitive disease [4]. ASCT is therefore considered the optimal treatment modality for younger and fitter patients able to tolerate transplant-related toxicities. However, existing studies supporting ASCT in SCNSL are largely limited by small cohorts and heterogenous populations [5].

In this study, we evaluated the survival outcomes of patients with SCNSL who underwent thiotepa-based ASCT as either upfront consolidation or treatment for relapsed or refractory disease at two national cancer centers in Singapore over the past decade.

Methods

Study design and patient selection

Patients with secondary CNSL who underwent thiotepa-based ASCT between 2013 and 2024 at the Singapore General Hospital and National University Hospital Singapore were included in this retrospective study (IRB: SGH 2015/2419, NUH 2010/00431). Patient demographics, disease status prior to transplant, treatment regimen and transplant details, and outcomes following ASCT were extracted from electronic records. Patients with primary CNSL were excluded from this study. Choice of treatment and decision for transplant were made based on physician and patient discretion guided by international hematopoietic cell transplantation guidelines [6].

Disease definitions

Patients with secondary CNSL who underwent ASCT were categorized into three groups: de novo systemic and CNS disease at diagnosis (de novo), previously treated disease with isolated CNS relapse (isolated CNS relapse), or previously treated disease with both systemic and CNS relapse (synchronous relapse). In this study, the histological diagnosis for all cases of SCNSL was determined to be large B cell lymphoma.

End point definitions

Progression-free survival (PFS) was defined as the time from date of ASCT to disease progression or death. Overall survival (OS) was defined as the time from date of ASCT to death from any cause. Lymphoma-related mortality was defined as death in the presence of lymphoma disease not in remission. Non-relapse mortality (NRM) was defined as death without evidence of disease progression. Relapses and progression of disease were considered competing risk for NRM.

Statistical analysis

For univariate analysis, chi-squared tests were used to compare categorical variables. Log-rank analysis was used to compare different groups. Two-sided p values of 0.05 were considered statistically significant. Survival probabilities, overall survival and progression-free survival were calculated using the Kaplan-Meier method. These analyses were performed using GraphPad Prism version 10.6.1 for Windows, GraphPad Software, Boston, Massachusetts USA, www.graphpad.com. For multivariate analysis, cox regression was performed using the R programming language (Vers. 4.3.3, R Core Team 2024) and the Survival package (Vers. 3.8-3, Thernaeau, 2024). Bootstrap analysis was used to ensure the robustness of our 4-variable multivariate model.

Results

Patient demographics

Our study included 27 patients with a median age of 57 years (range, 28–70) at time of ASCT. 15 (55.6%) were male and 12 (44.4%) were female. The majority (24/27, 88.9%) of these patients, had a performance status of ECOG 0–1 while 3 (11.1%) patients had ECOG 2–4. Five patients (18.5%) had de novo SCNSL, 14 patients (51.9%) had isolated CNS relapse SCNSL and 8 patients (29.6%) had synchronous relapse SCNSL. The median prior lines of treatment for those with relapsed SCNSL were 2. At the time of transplant, 16 patients (59.3%) were in complete remission (CR), 7 patients (25.9%) in partial remission (PR), and the remaining 4 patients (14.8%) had either stable or progressive disease (SD/PD). Anthracycline-based induction chemotherapy was most frequently used as first-line therapy, administered to 26 of 27 patients (96.3%). RDHAP (n = 11) and MATRix (n = 10) were more commonly used in subsequent lines of therapy. Thiotepa–carmustine–based conditioning chemotherapy was used in all but one patient (Table 1).

Table 1.

Study population baseline characteristics

Secondary CNSL		N = 27
Median age at transplant (years)		57 (28–70)
ECOG	0–1	24 (88.9%)
	2–4	3 (11.1%)
Sex	Male	15 (55.6%)
	Female	12 (44.4%)
Diagnosis at time of ASCT	De novo	5 (18.5%)
	Isolated CNS relapse	14 (51.9%)
	Synchronous relapse	8 (29.6%)
Disease status at time of ASCT	CR	16 (59.3%)
	PR	7 (25.9%)
	SD/PD	4 (14.8%)
Median lines of treatment prior to transplant		2 (1–5)
Treatment lines prior to ASCT		1st line	2 L+
	MATRix	0	10
	RMVP/RMVP-AraC +/- RT	1	1
	Anthracycline-based	26	2
	RDHAP	0	11
	WBRT	0	3
	Other (HiDAC/RICE/MTX-IFO-DEP)	0	4
Conditioning chemotherapy	TT/BCNU-based	26 (96.3%)
	TT/Busulfan/Cyclophosphamide	1 (3.7%)

CNSL Central Nervous System Lymphoma, ECOG Eastern Cooperative Oncology Group, ASCT Autologous stem cell transplant, CR Complete Remission, PR Partial Remission, SD Stable disease, PD Progressive Disease, 2 L+ 2nd line and beyond, MATRix Methotrexate/ Cytarabine/ Thiotepa/ Rituximab, RMVP Rituximab/ Methotrexate/ Vincristine/ Procarbazine, AraC Cytarabine, WBRT Whole Brain Radiotherapy, RDHAP Rituximab/ Dexamethasone/ High-dose Cytarabine/ Cisplatin, HiDAC High-dose Cytarabine, RICE Rituximab/ Ifosfamide/ Carboplatin/ Etoposide, MTX Methotrexate, IFO Ifosphamide, DEP Liposomal cytarabine, TT Thiotepa, BCNU Carmustine

Efficacy

With a median follow-up duration of 31 months (range, 8–50), among five patients with de novo secondary CNSL, the median PFS and OS were not reached. Of these five patients, two experienced early disease relapse at 1 and 4 months post ASCT and subsequently died. No patients in this group experienced NRM. The remaining three patients remained in remission following ASCT, ranging from 8 to 50 months post ASCT.

Further analyses focused on 22 patients with relapsed secondary CNSL. With a median follow-up of 63 months (range, 5–109), the median PFS and OS were 11.5 months and 30 months, respectively (Fig. 1A), corresponding to 4-year PFS and OS rates of 35.0% and 48.0%, respectively. Patients with isolated CNS relapse had a median PFS of 16 months and a median OS of 30 months, with 4-year PFS of 40.8% and OS of 38.7%. Patients with synchronous relapse had a median PFS of 5.5 months and a median OS that was not reached, with 4-year PFS of 25% and OS of 62.5%. Survival outcomes did not appear to differ significantly by subtype of relapsed secondary CNSL (PFS, p = 0.41; OS, p = 0.50) but was numerically superior in isolated CNS relapse. When stratified by sex, female patients demonstrated significantly superior outcomes, with a median PFS of 64 months compared to 5.5 months in males (p = 0.04), and a median OS that was not reached versus 11 months, respectively (p = 0.02) (Fig. 1B). This translated to superior 4-year PFS and OS of 50% and 70% in females, respectively, in comparison to 4-year PFS and OS of 25% and 27.8% in males, respectively. Stratification by disease status at the time of ASCT demonstrated that the 19 patients in CR/PR showed a median PFS and OS (16 months and not reached, respectively) compared with the three patients in SD/PD (3 months and 4 months, respectively), these differences were not statistically significant (PFS, p = 0.59; OS, p = 0.26). In the SD/PD group, one of 3 patients were progression-free at 4 years, resulting in a 4-year PFS of 33.3%, compared to the CR/PR group where 4 of 19 were progression-free, resulting in a 4-year PFS of 35.1%, (Fig. 1C). There was no significant difference in median PFS (p = 0.38) or OS (p = 0.81) between patients who received two versus more than two prior lines of therapy (Table 2) (Fig. 2).

Fig. 1.

Relapsed secondary CNSL PFS and OS data. PFS (left) and OS (right) for A all relapsed SCNSL cases, B Females versus Males, and C CR/PR versus SD/PD. *p-value<0.05. ns: not statistically significant

Table 2.

Univariate analysis of relapsed secondary CNSL outcomes

	n	Median PFS (months)	p-value	Median OS (months)	p-value	4-year PFS	4-year OS
All	22	11.5		30		35.0%	48.0%
Isolated CNS relapse	14	16	0.41	30	0.50	40.8%	38.7%
Synchronous relapse	8	5.5		NR		25.0%	62.5%
CR/PR	19	16	0.59	106	0.26	35.1%	50.2%
SD/PD	3	3		4		33.3%	33.3%
Female	10	64	0.04*	NR	0.02*	50.0%	70.0%
Male	12	5.5		11		25.0%	27.8%
ECOG 0–1	19	7	0.25	30	0.40	29.6%	44.7%
ECOG 2–4	3	NR		NR		66.7%	66.7%
2 treatment lines	17	7	0.38	106	0.81	41.2%	50.4%
> 2 treatment lines	5	18		30		0.0%	40.0%

PFS Progression-Free Survival, OS Overall Survival, * p-value<0.05

Fig. 2.

Relapsed SCNSL PFS and OS data for remaining variables. PFS (left) and OS (right) for A Isolated CNS relapsed SCNSL versus synchronous relapsed SCNSL, B ECOG 0-1 versus ECOG 2-4, and C: 2 lines versus more than 2 lines of treatment. *p-value<0.05. ns: not statistically significant

On multivariate analysis of relapsed SCNSL, after correcting for variables including subtype of relapsed SCNSL, disease status and number of prior treatment lines, female sex was the main predictor of improved PFS (HR 0.28, p = 0.04) and OS (0.15, p = 0.02) (Table 3). CR/PR disease status appeared to trend toward improved OS but this was not statistically significant (HR 5.01, p = 0.09). Note that our 4-variable model excluded ECOG status and age to ensure multivariate model stability, given that patients proceeding to autologous transplant are generally of good performance status by virtue of eligibility, and that age was non-contributory in all models.

Table 3.

Multivariate analysis of outcomes in relapsed secondary CNSL

Variables	PFS
	Hazards ratio	p-value	95% Confidence Interval
Relapsed SCNSL subtype: Isolated CNS vs. Synchronous	2.31	0.18	0.67 to 7.90
Sex: Female versus male	0.28	0.04*	0.08 to 0.96
Disease status at transplant: CR/PR versus SD/PD	4.41	0.11	0.73 to 26.60
Number of treatment lines: 2 lines versus more than 2 lines	2.78	0.16	0.67 to 11.60

	OS
Relapsed SCNSL subtype: Isolated CNS vs. Synchronous	0.69	0.59	0.17 to 2.74
Sex: Female versus male	0.15	0.02*	0.03 to 0.69
Disease status at transplant: CR/PR versus SD/PD	5.01	0.08	0.79 to 31.60
Number of treatment lines: 2 lines versus more than 2 lines	0.70	0.62	0.17 to 2.90

* p-value <0.05

Relapse and mortality

Out of 22 patients with relapsed secondary CNSL, 15 of them progressed with 12 deaths. Nine patients (40.9%) died from lymphoma-related mortality and 2 patients (9.1%) died from NRM (both male and due to pneumonia). Although not statistically significant, a trend toward higher relapse rate in males (83.3%) compared to females (50%) was observed. Subsequent therapy received after relapse post ASCT were summarized (Table S1).

Discussion

Thiotepa-based ASCT consolidation has been considered the standard of care for Primary CNSL but data of its use in Secondary CNSL remains controversial. The pivotal IELSG42 MARIETTA study demonstrated an overall 2-year PFS and OS of approximately 45% and dismal outcomes in the relapsed SCNSL group, particularly those with synchronous relapsed SCNSL [3]. In our study comprising of 22 cases of relapsed SCNSL over the past decade, the 4-year PFS and OS were similar at 35% and 48% respectively. Although our study reported a lower PFS rate than the IELSG42 MARIETTA trial in those with synchronous SCNSL, our cohort had a higher OS rate. This is possibly explained by the efficacy of subsequent salvage therapies available as shown in Table S1.

Based on both univariate and multivariate analysis, female sex was associated with an improved median PFS at 64 months compared to males at 5.5 months (p = 0.04), and an improved median OS that was not reached in females compared to 11 months (p = 0.02) in males. This is reminiscent of previous observations that male sex tended to do poorer in non-Hodgkin lymphomas [7], as seen in the post hoc analysis of the RICOVER-60 trial [8] that saw an improvement in the PFS but not OS by increasing the Rituximab dose in the elderly male cohort [9]. Although not definitively proven in those studies, this was theorized to be a result of differential pharmacokinetics leading to increased chemotherapy clearance in males compared to females. These findings should be interpreted with caution as we observed a difference in age and performance status where males had a higher median age of 57 years compared to females at 53.5 years, and that males had a median ECOG of 1 compared to females with a median ECOG of 0. It would be pertinent to mention that the counterintuitively longer survival outcomes for our ECOG 2–3 cohort is likely attributed to a minute sample size of n = 3.

Previous studies have identified complete remission and receipt of fewer than two prior lines of therapy as factors associated with improved PFS and OS⁵. Additionally, disease relapse in the leptomeningeal versus parenchymal compartments may also influence outcomes [10], along with consolidation with ASCT (PFS HR 0.36; OS HR 0.33). In this recent study involving the Mayo Clinic cohort, remission status and sustained treatment response exerted a significant influence on ASCT outcomes, highlighting the importance of chemosensitivity in consolidation strategies [10]. Although our study did not reach statistical significance in showing that achieving complete or partial remission (CR/PR) at the time of ASCT was a strong predictor of superior PFS and OS compared with stable or progressive disease (SD/PD), the benefit of ASCT is undoubtedly dependent on intrinsic chemosensitivity.

A large international retrospective study of SCNSL patients from the UK, US, and Canada identified ASCT as an intervention associated with improved survival outcomes [11], although this did not reach statistical significance in patients with synchronous relapsed disease. In that study, outcomes following ASCT were compared with CAR-T cell therapy using propensity score matching, with ASCT associated with superior survival. However, given the likely selection bias favouring less chemo-sensitive disease in patients receiving CAR-T therapy, these findings should be interpreted with caution. Careful patient selection therefore remains critical when choosing between ASCT and CAR-T therapy.

We did not observe a significant difference in outcomes based on the number of treatment lines prior to ASCT, likely confounded by the small number of patients who had received more than two prior lines of therapy (n = 5). Nonetheless, as cumulative salvage therapies may adversely impact patient fitness, alternative treatment modalities should be considered for patients who fail to respond to initial salvage chemotherapy, including anti-CD19 chimeric antigen receptor T-cell (CAR-T) therapy. CAR-T cell therapy has transformed the management of relapsed large B-cell lymphoma as an effective second-line [12, 13] and third-line [14] treatment option. Although pivotal registration trials of anti-CD19 CAR-T excluded patients with CNS involvement due to concerns regarding immune effector cell–associated neurotoxicity syndrome, emerging real-world data support its feasibility in CNS lymphoma [15]. Studies including both primary and secondary CNS lymphoma have reported 1- to 2-year PFS rates of 28–35% [16, 17], without evidence of increased neurotoxicity compared with patients without CNS disease.

Finally, while our analysis focused on relapsed SCNSL due to small numbers in the de novo SCNSL cohort (n = 5), 3 out of 5 patients with de novo SCNSL remained in remission post ASCT. This is consistent with studies reporting a 3-year PFS and OS of 74% and 80%, respectively, for patients with de novo SCNSL undergoing ASCT in first CR, versus a 3-year PFS and OS of 32% and 48% without ASCT consolidation [18], demonstrating efficacy of ASCT consolidation in this population.

Our study did come with several limitations. First, the small single-cohort sample size constrains statistical power and the stability of our multivariable estimates. The wide confidence intervals around several hazard ratios reflect this limitation, and the multivariable associations we report – including the apparent benefit associated with female sex and the adverse trend associated with stable and progressive disease at transplant – should therefore be interpreted as hypothesis-generating rather than confirmatory. Validation in larger cohorts is required before these findings can inform clinical practice. Second, the wide range of follow up among subgroups might affect interpretation of longer-term efficacy data, particularly in de novo SCNSL. However, when focusing on relapsed SCNSL, this group had a median follow-up duration of over 5 years (63 months) and show that a proportion of patients were able to obtain long term remission with thiotepa-based ASCT, with 4-year PFS of 35% and a low NRM at 9%.

In conclusion, thiotepa-based ASCT may provide durable disease control in patients with chemo-sensitive SCNSL and could be considered as a consolidation strategy for transplant-eligible patients. Future studies are needed to better define the role and optimal sequencing of CAR-T cell therapy, targeted agents, and immunotherapies within the treatment landscape of this aggressive disease.

Abbreviations

SCNSL

Secondary Central Nervous Lymphoma

ASCT

Autologous stem cell transplant

CR

Complete response

PR

Partial response

SD

Stable disease

PD

Progressive disease

PFS

Progression free survival

OS

Overall survival

CAR-T

Chimeric antigen receptor T cells

SGH

Singapore General Hospital

NUH

National University Hospital

NRM

Non-relapse mortality

ECOG

Eastern Cooperative Oncology Group

MATRix

Methotrexate/ Cytarabine/ Thiotepa/ Rituximab

RDHAP

Rituximab/ Dexamethasone/ High-dose Cytarabine/ Cisplatin

RMVP

Rituximab/ Methotrexate/ Vincristine/ Procarbazine

AraC

Cytarabine

WBRT

Whole Brain Radiation therapy

HiDAC

High-dose Cytarabine

RICE

Rituximab/ Ifosfamide/ Carboplatin/ Etoposide

MTX

Methotrexate

IFO

Ifosphamide

DEP

Liposomal cytarabine

TT

Thiotepa

BCNU

Carmustine

NR

Not reached

NS

Not statistically significant

BTKi

Bruton Tyrosine Kinase inhibitor

Authors’ contributions

LN and MP contributed to the conception and design of the study and are co-last authors. MWT, VL, JYT and LN curated the data. MWT, JYT and LN processed and analysed the data. MWT wrote the manuscript under the supervision of LN. All authors read and approved the final manuscript.

Funding

No funding has been received for the purposes of this study.

Data availability

All data supporting the findings of this study are available within the paper and its Supplementary Information. These data are available upon request.

Declarations

Ethics approval and consent to participate

Consent for data collection for all participants were obtained and approved as part of the SingHealth Centralized Institutional Review Board (CIRB) 2015/2419 and National University Health System (NUHS) Ethics and Compliance Online System (ECOS) 2010/00431. This research study was conducted in accordance with guidelines and regulations stipulated by the IRB above, and with the Declaration of Helsinki. The requirement for informed consent was waived by the relevant ethics committee due to the retrospective nature of the study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.