International multicenter retrospective analysis of thiotepa-based autologous stem cell transplantation for secondary central nervous system lymphoma

Secondary central nervous system lymphoma (SCNSL) is a rare, aggressive disorder with a historically dismal prognosis of <6 months.¹ Patients may present de novo with systemic disease or at relapse, either with isolated central nervous system (CNS) disease or synchronous systemic involvement. These differing presentations create the therapeutic challenge of controlling both the systemic and CNS disease. Thiotepa-based autologous stem cell transplant (ASCT) in first remission has been explored in SCNSL as a means of overcoming the poor outlook. Retrospective studies including consolidative ASCT in SCNSL generally include small series of patients with heterogeneous histological subtypes. Transplant-specific outcomes are not well characterized.^2-4 Performing large trials is challenging, with the largest prospective series reporting only 37 patients proceeding to ASCT.⁵ The largest retrospective series (n=151) reported no patients who had received thiotepa-based conditioning, with the majority having undergone BEAM (carmustine, etoposide, cytarabine, melphalan)-conditioned ASCT.⁶ Thiotepa-based conditioning with carmustine or busulfan has greater CNS bioavailability⁷ compared with BEAM and produces superior outcomes in primary CNS lymphoma.⁸ We analyzed the survival outcomes of the largest cohort of patients with SCNSL, focused exclusively on patients with diffuse large B-cell lymphoma (DLBCL) or transformed lymphoma, who were treated with chemoimmunotherapy and consolidated with thiotepa-conditioned ASCT.

Consecutive adult patients treated from January 31, 2013 to February 24, 2020 across 17 centers and three countries (UK, Italy and Germany) with thiotepa-based ASCT consolidation were retrospectively reviewed. Patients were followed up to December 1, 2021. CNS involvement was confirmed by brain biopsy and/or cerebrospinal fluid studies and/or neuroimaging. Baseline characteristics, details of therapy and response were collected. The primary endpoints were 3-year progression-free survival (PFS) and overall survival (OS) from time of stem cell infusion; secondary endpoints were the incidences of CNS and systemic relapse and of non-relapse mortality (NRM). OS and PFS estimates were generated using the Kaplan-Meier method and groups were compared using Cox regression and the log-rank test. Backwards selection with P=0.05 for inclusion was used for multivariable analyses. All statistical analyses were conducted using STATA v16.1 software (STATA corp, College Station, TX, USA).

One hundred thirty-four patients (85 male, 49 female) with SCNSL underwent thiotepa-conditioned ASCT. These patients’ baseline characteristics are outlined in Table 1. Forty-four patients did not have a CNS biopsy and were diagnosed based on a biopsy from a systemic site or neuroimaging alone. At the time of SCNSL diagnosis, 52 (39%) patients had a de novo presentation of SCNSL (synchronous systemic and CNS disease and were treatment-naïve) and 82 (62%) patients had relapsed diffuse large B-cell lymphoma, of whom 62 (46%) had isolated CNS relapse and 20 (15%) had a synchronous relapse presentation (systemic and CNS disease with prior therapy). For those with CNS involvement at relapse, the majority (77/82; 94%) had received prior chemotherapy with R-CHOP (rituximab plus cyclophosphamide, doxorubicin, vincristine and prednisone), including two patients who were given etoposide in addition to R-CHOP. Among all patients, methotrexate-cytarabine-based induction was most frequently used (n=123; 92%). Complete responses or partial responses to induction, as assessed before ASCT by positron emission tomography (PET) with computed tomography (CT) or CT alone, were achieved in 77/94 (82%) and 13/94 (14%) patients, respectively, and by 83/127 (65%) and 37/127 (29%), respectively, according to magnetic resonance imaging (MRI) of the head. The conditioning regimens employed were carmustine-thiotepa (n=112; 84%), busulfan-thiotepa (n=18; 13%), busulfan-lomustinethiotepa (n=2; 1%), thiotepa-etoposide-cytarabine-melphalan (n=1; 1%) and thiotepa alone (n=1; 1%). The median number of CD34⁺ cells infused was 4.4x10⁶/kg (range, 1.4x10⁶/kg - 37.1x10⁶/kg). The median days to neutrophil and platelet engraftment were 11 (interquartile range, 10-12) and 13 (interquartile range, 11-17). Neutrophil and platelet engraftment were defined as the first of 2 consecutive days with an absolute neutrophil count >0.5x10⁹/L and a platelet count >20x10⁹/L, without transfusion support.

At ASCT, the median duration of hospitalization was 22 days (range, 14-298) and the Intensive Care Unit admission rate was 8% (11/130). Grade 3-4 renal impairment was observed in 6% (8/130) and hepatic impairment in 4% (5/130). With a median follow-up of 47 months (interquartile range, 29-60), the 3-year OS and PFS rates were 71.6% (95% confidence interval [95% CI]: 61.9% – not reached) and 61.1% (95% CI: 52.2-68.9%), respectively (Figure 1). Ninety patients with histologically confirmed CNS disease and 44 patients assessed with neuroimaging alone had similar OS (3-year rates: 70.2% [95% CI: 59.3-78.7] vs. 67.2% [95% CI: 50.9-79.1], log rank P=0.92) and PFS (3-year rates: 59.0% [95% CI: 47.9-68.5] vs. 65.5% [95% CI: 49.4-77.6], P=0.44). During the study period, 48 patients died, 43 relapsed and 14 died without documented relapse. The 100-day NRM was 3% and the cumulative incidence at 1 and 3 years was 8.4% (4.7-14.6). Causes of NRM were infection (6/14), respiratory failure (2/14), secondary acute myeloid leukemia (1/14) and unknown (5/14: all beyond day 100). Most relapses occurred within 2 years of ASCT (34/43; 79%).

Table 1.

Patients’ baseline characteristics prior to autologous stem cell transplantation.

The optimal depth of disease response that must be achieved prior to ASCT has previously been uncertain. Our data indicate that patients with a partial response (CNS, systemic or both) prior to ASCT have good outcomes. Those with a partial response after induction chemotherapy in the systemic compartment (by PET-CT/CT) or in the CNS (by MRI) did not differ significantly for PFS, OS or time to relapse when compared with those who had a complete response (Table 2, Online Supplementary Table S1). Combining response data showed a better OS for patients who were in complete remission according to both PET and MRI than for those in partial remission according to either technique (P=0.032, P=0.076, and P=0.055). Two of six patients transplanted with progressive disease responded, and are in complete remission; nevertheless outcomes were worse than those in all other patients, with four of the six progressing.

Adverse predictors of PFS and OS on univariable analysis were older age, Eastern Cooperative Oncology Group score 2-3, number of prior lines of therapy for SCNSL and progressive disease on pre-ASCT MRI. Presentation (relapsed DLBCL with synchronous presentation vs. de novo/isolated relapse) was significantly associated with inferior PFS. The only factors that were associated with poorer PFS in multivariable analysis were synchronous presentation, age and two or more prior lines of therapy. For OS, only age and two or more lines of SCNSL treatment remained significant. This is consistent with data in primary CNS lymphoma and systemic DLBCL.⁹

Patients presenting with synchronous relapse of SCNSL remain a challenge and have the poorest outcomes. The 3-year PFS in this group was 40.0% (19.3-60.1), compared to 62.7% (47.9-74.4) and 67.7% (53.1-77.1) in the groups with de novo and isolated relapse presentations (Table 2). This is comparable to the CORAL data of a 3-year PFS of 39% in 68 patients with relapsed/refractory DLBCL undergoing BEAM-conditioned ASCT.¹⁰ In our cohort this appears to be driven by a higher rate of systemic relapse after ASCT (55.0% vs. 6.0% de novo vs. 2.1% isolated) and may therefore reflect the difficulty in achieving control of systemic disease at relapse. The risk of systemic failure was greater for those with a synchronous relapse presentation than those with de novo/isolated presentations (hazard ratio synchronous vs. de novo = 14.36 [95% CI: 4.03-51.1%], hazard ratio synchronous vs. isolated = 54.64 [95% CI: 7.1-421.8], log rank P<0.0001).

Relapse after ASCT resulted in very poor outcomes. As in the CORAL study, a shorter time to relapse after ASCT was associated with inferior survival.¹⁰ In our study, 43 patients relapsed after ASCT (27 CNS only, 13 systemic only, 3 both), at a median of 4.9 months (range, 1-49.3); 34 died with a median survival of 3.7 months (range, 2.1-7.2). Those relapsing <3 months after ASCT had a median survival of 1.5 months (95% CI: 0.72-2.04) compared with 3.7 months (95% CI: 3.01-4.37) for those who relapsed 3-6 months after ASCT and 21.6 months (95% CI: 9.6-not reached) for those who relapsed at ≥6 months (log rank P<0.0001). Of 21 patients receiving salvage chemotherapy, 15 (71%) have died, all due to progressive disease.

Overall, our data support thiotepa-based ASCT as a standard of care of conditioning in SCNSL. Our data suggest that patients with SCNSL undergoing this strategy have superior OS and PFS compared to cohorts receiving BEAM conditioning, although the proportion of SCNSL presentation was not characterized in these studies.^4,6 No patients underwent thiotepa-busulfan-cyclophosphamide conditioning which has been used in primary CNS lymphoma with higher rates of NRM and a similar risk of all-cause mortality after 6 months. In our study, the 100-day NRM was 3% and 8.4% at 3 years, with others reporting 100-day NRM of approximately 10% in SCNSL.^2,3 Hematopoietic recovery times and intensive care admission rates were comparable to those previously published.

Factors significantly associated with inferior PFS and OS in our series included number of prior lines of therapy for SCNSL and older age. Despite this, carefully selected patients >70 years still have good outcomes and should not be excluded. Two prospective trials included patients ≤70 years old, with restrictive criteria for organ function and exclusion of those with human immunodeficiency virus infection or hepatitis.^5,11 There are no prospective data for patients >70 years old.^12,13 Our unselected retrospective series reflects real-world practice: 30% (38/127) would not have met MARIETTA trial eligibility criteria at SCNSL diagnosis (n=30) or prior to ASCT (n=8) (age up to 77 years [>70 years, n=17; 13%] at SCNSL diagnosis, prior high-dose methotrexate use [n=13; 10%], well-controlled human immunodeficiency virus infection [n=2; 1%], impaired renal function prior to ASCT [glomerular filtration rate <60 mL/min, 6/129; 5%] and left ventricular ejection fraction <50% [3/112; 3%]).

Figure 1.

Outcomes afer autologous stem cell transplantation for secondary central nervous system lymphoma. (A) Progression-free survival. (B) Overall survival. (C) Incidence of systemic relapse after autologous stem cell transplantation (ASCT). (D) Incidence of isolated central nervous system relapse after ASCT.

Our data are retrospective and have inherent limitations. We were unable to accurately identify all patients presenting with SCNSL and only included those who proceeded to ASCT. Forty-four percent of those with a relapsed SCNSL presentation presented within a year of a diagnosis of DLBCL, whereas typically 90% of CNS relapses occur during the first year of follow-up,⁵ demonstrating a possible selection bias as we postulate a cohort of patients who relapse early may not proceed to ASCT. Data were incomplete or not uniformly performed on baseline risk factors (including cell of origin/gene rearrangements) and therefore analysis of potential confounders may be limited. Despite this being the largest cohort of SCNSL patients treated with thiotepa-conditioned ASCT to date, good outcomes (therefore small numbers of events) limited our ability to run full multivariable models or multivariable analysis by relapse type, and treatment choice bias will limit any comparison of treatment regimens.

In conclusion, thiotepa-conditioned ASCT is an effective consolidation therapy with low NRM and leads to durable responses particularly in those with de novo or isolated relapse presentation. Advanced age (>70 years) does not preclude consideration for this consolidation strategy. Patients presenting with synchronous SCNSL at relapse have poor outcomes, mainly due to post-ASCT systemic relapse, and may benefit from a different treatment approach. Patients having a partial or complete response after induction therapy can achieve durable remissions with thiotepa-based ASCT. The lack of requirement of a complete response prior to ASCT may help to minimize treatment-related toxicity by shortening courses of induction chemotherapy.

Table 2.

Risk factors for progression-free survival and overall survival.