Multicenter Randomized Phase II Study of R-MPV-A Chemoimmunotherapy With or Without Low-Dose Whole-Brain Radiotherapy for Newly-Diagnosed Primary CNS Lymphoma

Abstract

Background:

Methotrexate-based chemoradiotherapy is effective in primary central nervous system lymphoma (PCNSL) but carries a risk of significant neurotoxicity, especially in the elderly. In a single-arm study, a regimen with methotrexate, procarbazine, vincristine and cytarabine was combined with rituximab (R-MPV-A) and substantially reduced doses of whole-brain radiotherapy (LD-WBRT), resulting in excellent progression-free survival (PFS) and overall survival (OS). Because R-MPV-A had never been tested without radiation, we sought to evaluate the efficacy of R-MPV-A with and without LD-WBRT, as well as determining if such low radiotherapy doses influenced disease control and/or neurotoxicity.

Methods:

Patients were randomized to receive R-MPV-A alone (Chemo arm) or combined with LD-WBRT (ChemoRT arm), given at 23.4 Gy (1.8 Gy X 13). Primary endpoint was intent-to-treat (ITT) PFS. A sample size of 89 would provide 80% power to detect a hazard ratio (HR) of 0.63 (one-sided alpha=0.15).

Results:

Ninety-one patients were randomized, with 44 analyzed in the ChemoRT and 46 in the Chemo arm. Median age was 66 and 59.5, respectively. R-MPV-A was well tolerated, achieving complete response rate of 92.3% (ChemoRT) and 76.3% (Chemo). After median follow-up of 4.6 years, median PFS was not reached (ChemoRT) vs 2.1 years (Chemo), HR=0.47 (p=0.007; 95% CI: 0.26 to 0.87). The 2-year PFS was 78.7% vs 54%, respectively. Although the HR for OS was an impressive 0.71, this did not reach statistical significance (p=0.33), given the limited number of events. Importantly, no inferiority in neuropsychological evaluation for the ChemoRT arm was observed, and in fact, several tests favored the ChemoRT arm.

Conclusions:

R-MPV-A is a highly efficacious and safe regimen with or without LD-WBRT. LD-WBRT contributes to disease control, increases PFS in PCNSL, could potentially have an impact on OS, and does not increase cognitive decline.

Introduction

Primary CNS lymphoma is an aggressive non-Hodgkin lymphoma restricted to the brain, eyes, spinal cord, and cerebrospinal fluid. In immunocompetent patients, features of a diffuse large B cell lymphoma (DLBCL) and CD20 expression are observed in more than 95% of tumors.

Historically, improved outcomes in this disease were first observed with whole-brain radiotherapy (WBRT), initially utilized at doses of 45 Gy or higher^1,2. The first attempts to integrate chemotherapy were unsuccessful, possibly reflecting the poor penetration of CHOP and other standard regimens utilized in systemic DLBCL^3,4. When high-dose methotrexate-based regimens with better CNS penetration were added to radiation, remarkable improvements in progression-free survival (PFS) and overall survival (OS) were observed^5–8. However, late delayed neurotoxicity emerged as a concern in long-term survivors, often associated with devastating if not lethal consequences, especially in the elderly^9,10. Consequently, in spite of the established efficacy of WBRT, attempts were made to eliminate it using more intense chemotherapy, requiring rescue with stem cell transplant, an aggressive regimen that many elderly and sick patients would either not qualify for or not tolerate well.

Therefore, in order to determine whether an effective modality, i.e. WBRT could be safely integrated in the management of this disease without resorting to dose-intense chemotherapy requiring autologous stem cell transplant, we previously conducted a phase 2 single-arm study utilizing a chemotherapy regimen with methotrexate, procarbazine, vincristine and cytarabine combined with rituximab (R-MPV-A) and substantially reduced doses of WBRT (1.8 Gy per fraction to a total of 23.4 Gy)^11,12. In that single center trial, excellent results were observed in the patients who proceeded with WBRT, with median PFS of 7.7 years and 5-year OS of 80%. Results were particularly striking in the elderly population, where a median PFS of 4.4 years was achieved, well above historical controls. Importantly, neurocognitive evaluation showed improvements in cognition following initial treatment, with no significant decline observed throughout the follow-up period¹¹. However, interpreting those results was challenging. The R-MPV-A regimen had never been utilized without radiation or myeloablative consolidation chemotherapy, and it was therefore unclear if the low dose of radiotherapy was of sufficient biological activity to improve disease control, and how the observed neurotoxicity profile would compare to outcomes with R-MPV-A alone, especially in a multicenter setting, where a more representative patient population could be accrued. In this randomized study, we sought to investigate these important questions by evaluating the efficacy and safety of R-MPV-A with and without consolidation low-dose WBRT (LD-WBRT) in newly-diagnosed primary CNS lymphoma, identify any potential OS benefit signal, and evaluate cognitive outcomes.

Patients and methods

This was a randomized, prospective, multicenter phase II study sponsored by the National Institutes of Health/ National Cancer Institute/ Cancer Therapy Evaluation Program, through the former Radiation Therapy Oncology Group (RTOG), now NRG Oncology cooperative group (study ID: RTOG 1114; NCI-2011–02678; clinicaltrials.gov: NCT01399372). The study was approved by the Institutional Review Board of participating institutions. Written informed consent was obtained from all patients or guardians. Eligible patients were registered, stratified by MSKCC-RPA class¹³ and randomized to receive R-MPV-A with LD-WBRT (ChemoRT arm) or R-MPV-A alone (Chemo arm). The study schema is shown in Figure 1. Additional results are provided in Appendix 1 (online) and further details on methods including full eligibility criteria, pre-treatment evaluations, treatment protocol and dose modification recommendations are provided in Appendix 2 (online).

Figure 1:

Study Schema

Key eligibility criteria included: newly-diagnosed B-cell non-Hodgkin’s lymphoma involving the brain; no history of immunodeficiency; adequate organ function; creatinine clearance >50 cc/min/1.73m²; Karnofsky performance status (KPS) ≥ 50, with the exception of patients with KPS 30 to 50, who were eligible if the reason for the poor performance status was neurologic deficit from disease (which would be expected to resolve quickly with onset of therapy). Patients could not have evidence of systemic non-Hodgkin lymphoma on a CT of the chest, abdomen and pelvis. Bone marrow biopsy was not required for registration but was obtained prior to start of treatment; patients were to be deemed ineligible if results demonstrated involvement by lymphoma.

Baseline staging included brain MRI, lumbar puncture, complete ophthalmologic evaluation including slit lamp and bone marrow biopsy. Hepatitis B screening was obtained; hepatitis B reactivation prophylaxis with entecavir 0.5 mg PO daily was recommended for patients with positive HB surface antigen or HB core antibody.

Treatment

For both arms, R-MPV was given for upto four cycles (1 cycle=28 days), as follows: Rituximab on days 1 and 15 at 500mg/m² IV; methotrexate 3.5 g/m² in 2-hour IV infusions days 2 and 16; procarbazine 100mg/m² PO days 2–8; vincristine 1.4 g/m² IV days 2 and 16 (dose capped at 2.4 mg and discontinued for cycles 3 and 4); and filgrastim 5 mcg/kg/day days 10 to 14 and 18–22. Methotrexate was administered as inpatient, as detailed in appendix 2.

Following R-MPV induction chemotherapy, patients in the chemoRT arm received LD-WBRT, followed by consolidation IV high-dose cytarabine. Patients in the chemo arm proceeded directly to high-dose cytarabine. Cytarabine was given in 2 cycles (1 cycle=28 days), at a dose of 3g/m²/day over 3 hours on days 1 and 2 of each cycle; pegfilgrastim was given at 6 mg SC on day 4 of each cycle.

Patients in the chemoRT arm received LD-WBRT to a total dose of 23.4 Gy (1.80 Gy per fraction X13 administered daily, upto 5 fractions per week). The opposed lateral radiation fields included the whole brain down to the level of C2 (“German helmet” shape) and excluded the anterior two-thirds of the orbit. Patients in the chemoRT arm with ocular involvement were irradiated without orbital shielding to the full dose of 23.40 Gy. If no ocular involvement was evident on initial examination then only the posterior one third of the orbit was to be included in the treatment portal (Appendix 2).

Follow-up

Disease status was assessed as per the International PCNSL Collaborative Group (IPCG) guidelines¹⁴ (appendix 2) at baseline, then every 2 months for two years, and every 6 months for three years, for a total of 5 years. Patients with progressive disease discontinued the study treatment and were offered salvage treatment at investigator’s discretion but remained in active follow-up.

Statistical Considerations, Randomization and Masking

Patients were stratified by the MSKCC RPA classification for PCNSL based on age and KPS (Class 1: age ≤ 50; Class 2: age > 50 and KPS ≥ 70; Class 3: age > 50 and KPS < 70). The randomization occurred in a 1:1 ratio, in a permuted block design, non-blinded.

The primary endpoint was intent-to-treat progression-free survival (PFS), defined as time from randomization until disease progression (as determined by local investigator) or death, whichever occured first. Secondary endpoints were overall survival (OS), objective response rate (ORR)¹⁴, acute toxicity, and evaluation of neurocognitive function/ treatment-related neurotoxicity and quality of life.

The trial was powered to detect an increase in median PFS from 12 months (chemoimmunotherapy alone) to 19 months for chemoimmunotherapy combined with LD-WBRT, translating to a hazard ratio (HR) of 0.63. With 1:1 randomization, a total of 84 eligible patients would yield ~80% power to detect a 37% hazard reduction using a one-sided 0.15 level log-rank test. Accounting for 5% ineligibility, target accrual was n=89. The Kaplan-Meier method and Cox proportional hazards models were utilized. Analyses were performed using SAS version 9.4. All secondary endpoints were tested using 2-sided 0.05 significance level; no multiple comparison adjustment was made.

The primary efficacy analysis was to be performed after 67 PFS events, with an interim futility analysis after 34 events. In July 2017, the NRG Oncology Data Monitoring Committee recommended an interim efficacy analysis after 67% of required events. If the estimated HR was <0.5 favoring ChemoRT, the trial result could be reported early, based on promising efficacy signal.

Neurocognitive Function and Quality of Life

Comprehensive neurocognitive testing and quality of life assessment were performed in consenting patients at baseline, as well as following conclusion of induction chemotherapy, and every 6 months thereafter. Patients with progressive disease at any time point continued to be followed with neurocognitive testing and quality of life assessments, except during active disease recurrence and salvage treatments, in which case evaluations were postponed and resumed after salvage treatment completion, following the original schedule.

Neurocognitive testing consisted of a modified IPCG battery¹⁵ testing verbal memory (Hopkins Verbal Learning Test-Revised [HVLT-R]), processing speed (Trail Making Test, Part A), and executive function (Trail Making Test, Part B; Controlled Oral Word Association [COWA]). Quality of life was evaluated utilizing the EORTC QLQ-C30/BCM20¹⁶. Detailed statistical analysis methodology is provided in Appendix 2.

In addition to the formal neurocognitive testing performed in patients who agreed to participate, the overall population of patients was evaluated for the development of neurotoxicity as deemed by the local investigators in routine clinical assessments. Specific consent was not necessary for this evaluation and all patients were included. To account for the competing risk of death, the cumulative incidence method and the Gray’s test were used to analyze time to neurotoxicity.

An ad-hoc analysis was performed to examine the association between disease progression and subsequent neurocognitive scores. In this analysis, a longitudinal model is fitted for each neuropsychological test with progression status as a time-varying covariate, while adjusting for baseline covariates including MSKCC RPA class, neurologic function, treatment, time, and a time by treatment interaction.

Data sharing

All data will be made available per the National Cancer Institute’s National Clinical Trials Network (NCTN) Data Archive rules. The link for the archive is: https://nctn-data-archive.nci.nih.gov/.

Results

Patient characteristics

Accrual was completed as planned. From September 2011 to January 2016, a total of 91 patients were randomized across 26 institutions (Figure 2, CONSORT diagram). One patient in the Chemo arm was deemed ineligible due to bone marrow involvement by lymphoma. A total of 44 eligible patients were enrolled in the ChemoRT arm and 46 in the Chemo arm. Median age was 63 years (min-max: 21–84), 47.8% were male, 81.1% were white, 35.6% had KPS 90–100, 47.8% had moderate or severe neurological symptoms. Histology was diffuse large B cell lymphoma in 95.6%. Patient characteristics were in general well-balanced between arms (Table 1). Out of 86 cases reported by sites as DLBCL, 67 had adequate materials for central pathology review. Among these 67 cases, 64 cases (96%) were confirmed as DLBCL per central review.

Figure 2:

CONSORT diagram

Table 1-.

Patient characteristics

		Chemo (n=46)	ChemoRT (n=44)	Total (n=90)
Age, years	Median (min-max)	60 (21 – 84)	66 (40 – 78)	63 (21 – 84)
	≤ 49	6 ( 13.0%)	7 ( 15.9%)	13 ( 14.4%)
	50 – 59	17 ( 37.0%)	9 ( 20.5%)	26 ( 28.9%)
	60 – 69	14 ( 30.4%)	14 ( 31.8%)	28 ( 31.1%)
	≥ 70	9 ( 19.6%)	14 ( 31.8%)	23 ( 25.6%)
Sex	Male	23 ( 50.0%)	20 ( 45.5%)	43 ( 47.8%)
	Female	23 ( 50.0%)	24 ( 54.5%)	47 ( 52.2%)
Race	Asian	6 ( 13.0%)	2 ( 4.5%)	8 ( 8.9%)
	White	35 ( 76.1%)	38 ( 86.4%)	73 ( 81.1%)
	Unknown/ not reported	5 ( 10.9%)	4 ( 9.1%)	9 ( 10.0%)
Ethnicity	Hispanic or Latino	1 ( 2.2%)	1 ( 2.3%)	2 ( 2.2%)
	Not Hispanic or Latino	40 ( 87.0%)	38 ( 86.4%)	78 ( 86.7%)
	Unknown/ not reported	5 ( 10.9%)	5 ( 11.4%)	10 ( 11.1%)
KPS	Median (min-max)	80 (30 – 100)	75 (40 −100)	80 (30 – 100)
	30–50	6 ( 13.0%)	6 ( 13.6%)	12 ( 13.3%)
	60–80	23 ( 50.0%)	23 ( 52.3%)	46 ( 51.1%)
	90–100	17 ( 37.0%)	15 ( 34.1%)	32 ( 35.6%)
Neurologic Function/Symptoms
	None	8 ( 17.4%)	7 ( 15.9%)	15 ( 16.7%)
	Minor	16 ( 34.8%)	16 ( 36.4%)	32 ( 35.6%)
	Moderate	18 ( 39.1%)	15 ( 34.1%)	33 ( 36.7%)
	Severe	4 ( 8.7%)	6 ( 13.6%)	10 ( 11.1%)
Source of Histologic Confirmation
	CSF cytology	6 ( 13.0%)	2 ( 4.5%)	8 ( 8.9%)
	Biopsy of vitreous or uvea	2 ( 4.3%)	1 ( 2.3%)	3 ( 3.3%)
	Brain biopsy	38 ( 82.6%)	41 ( 93.2%)	79 ( 87.8%)
Histology lymphoma	Diffuse large B cell	44 ( 95.7%)	42 ( 95.5%)	86 ( 95.6%)
	Lymphoblastic	0 ( 0.0%)	1 ( 2.3%)	1 ( 1.1%)
	Unknown	1 ( 2.2%)	1 ( 2.3%)	2 ( 2.2%)
	Other	1 ( 2.2%)	0 ( 0.0%)	1 ( 1.1%)
MSKCC RPA
	Class 1	7 ( 15.2%)	6 ( 13.6%)	13 ( 14.4%)
	Class 2	27 ( 58.7%)	25 ( 56.8%)	52 ( 57.8%)
	Class 3	12 ( 26.1%)	13 ( 29.5%)	25 ( 27.8%)
Ocular Involvement		4 (8.7%)	0 (0.0%)	4 (4.4%)
CSF Involvement		6 (13.0%)	6 (13.6%)	12 (13.3%)

KPS- Karnofsky Performance Status; MSKCC RPA- Memorial Sloan Kettering Cancer Center Recursive Partitioning Analysis; CSF- Cerebrospinal fluid

Efficacy

Primary Endpoint

At the time of this analysis, 44 PFS events (67% information) had occurred. The median PFS was not reached (ChemoRT arm) vs 2.1 years (Chemo); HR=0.47 (95% CI: 0.26 to 0.87), one-sided log rank p=0.007 (Figures 3 and 4). The 2-year PFS was 78.7% (ChemoRT [95% CI: 66.4–91.1%]) vs 54% (Chemo [95% CI: 39.1–69%]). Median follow-up for PFS was 4.6 years (95% CI: 4.1–5.1). Adjusting for MSKCC RPA class, neurologic function and sex, ChemoRT was associated with a 64% reduction in hazard compared to Chemo (HR=0.36, 95% CI: 0.19–0.70; p=0.002 [Appendix 1A]).

Figure 3:

Progression-free survival

Figure 4:

Exploratory analysis of progression-free surviva; (PFS) and overall survival by age

Secondary and Exploratory Endpoints

Response Rate and Overall Survival

A total of 77 evaluable patients had images for the blinded central radiologic reviews. Overall best response rate in the ChemoRT vs Chemo were 100% and 86.9% respectively (Fisher’s exact test p=0.03). In the Chemo RT arm, best response rates were complete response (CR) in 92.3%, CR unconfirmed (CRu) in 2.6%, PR in 5.1%, SD in 0%, and PD 0%. In the Chemo arm, best response rates were CR (76.3%), CRu (5.3%), PR (5.3%), SD (2.6%), and PD (10.5%). Therefore, the confirmed and total CR rates (including unconfirmed CR) were 92.3 and 94.7% for ChemoRT versus 76.3 and 81.4% for Chemo alone, representing an absolute increase in confirmed CR rates of 16%, as well as a 24% absolute increase in overall response rates in favor of ChemoRT (100% vs. 86%).

OS data were updated through August 4, 2021. The median follow-up for OS was 5.6 years [95% CI: 5.3–5.7]. The number of deaths was 14 and 18 in the ChemoRT and Chemo arm, respectively. The median OS was not reached in either arm, with HR=0.71 (95% CI: 0.35 to 1.42); two-sided log-rank p=0.33 (Figure 4). The estimated 5-year survival rates are 65.8% (95% CI: 51.2%−80.4%) in the ChemoRT and 58.7% (95% CI: 44%−73.4%) in the Chemo arm. Adjusting for MSKCC-RPA class, neurologic function and sex, HR for OS was 0.57 (95% CI: 0.27–1.20; p=0.14), Appendix 1B.

Acute toxicity

The immunochemotherapy regimen was relatively well tolerated in both arms. Acute toxicities were in general transient and reversible, consistent with previously described profile for R-MPV. Grades 3–5 toxicities deemed definitely, probably or possibly related to treatment are summarized in Appendix 1C. One patient died in the ChemoRT arm of an unknown cause; this death was conservatively deemed possibly related to treatment. One patient died in the Chemo arm from sepsis; the death was deemed unlikely related by the investigator but occurred 22 days after end of treatment.

Salvage treatments

Salvage treatments were utilized in 16 (Chemo arm) and 4 patients (ChemoRT). In the Chemo arm, treatments consisted of WBRT in 7 patients (23.4 Gy in 1 patient, 30 Gy or higher in 6); other types of radiotherapy in 6 (orbital n=3; focal RT n=2; stereotactic radiosurgery n=1); and chemotherapy in 15 (including regimens containing MTX n=9; rituximab n=9; ibrutinib n=5; pomalidomide n=1; liposomal cytarabine n=2; PI3K inhibitor n=2 and other cytotoxic regimens n=6). In the ChemoRT arm, one patient received RT to the spine, two patients received salvage regimens containing rituximab and MTX and one rituximab with temozolomide. High-dose chemotherapy with stem cell transplant (HDCASCT) has been given to 3 patients (Chemo arm) and 1 patient (ChemoRT arm).

Neurocognitive function, quality of life and neurotoxicity

A total of 72 (80%) patients agreed to participate in the neuropsychological testing (n=37 ChemoRT and n=35 Chemo arm). Of those, 32 patients (86.5%) in the ChemoRT and 30 (85.7%) in the Chemo arm completed baseline testing, with comparable patient characteristics [data not shown].

The cumulative incidence curves for time-to-neurocognitive failure are provided in Figure 5. At three years, the estimated cumulative incidence was 17.9% (95% CI: 6.3–34.1%) in the ChemoRT vs 41.6% (95% CI: 21–61.2%) in the Chemo arm. There was no statistically significant difference in time-to-neurocognitive failure between the arms (HR = 0.52, 95% CI: 0.21–1.31; Gray’s test: p=0.18).

Figure 5:

Time to neurocognitive failure

NOTE: Time to neurocognitive failure is defined as time to first cognitive failure on two or more of the following tests: HVLT-R Total Recall, Delayed Recall and Delayed Recognition, the COWA, Trail Making Test Part A or B.

In the longitudinal analysis (appendix 1D), there was a statistically significant improvement over time in the ChemoRT arm as compared to the Chemo arm in HVLT-R Delayed Recall (time by treatment interaction: p=0.001) and in HVLT-R Delayed Recognition (p < 0.001), after adjusting for MSKCC-RPA class and neurologic function status. Changes in the other scores were not statistically significantly different between arms.

A post hoc analysis examining the association between disease progression and subsequent neurocognitive failure showed that progression and salvage therapy were associated with worse standardized HVLT-R Total Recall, standardized HVLT-R Delayed Recall, standardized HVLT-R Delayed Recognition, standardized COWA, and the Clinical Trial Battery composite score (p<0.001).

In addition to the formal neurocognitive testing performed in patients who agreed to participate, the overall population of patients was evaluated for the development of neurotoxicity as deemed by the local investigators in routine clinical assessments. For this purpose, neurotoxicity was defined as clinical development of severe neurologic symptoms (cognitive deterioration, gait ataxia and psychomotor slowness), as compared to baseline post-treatment evaluation. Based on this assessment, the estimated cumulative incidence of neurotoxiticy at 3 years was 11.9% (95% CI: 4.3–23.8%) in the ChemoRT and 9.2% (95% CI: 2.9–20.1%) in the Chemo arm, with no statistically significant differences (HR=1.45, 95% CI: 0.47–4.51; Gray’s test: p=0.53).

Quality of life analysis (appendix 1E) showed similar trends as observed in the neurocognitive testing. There was a significantly greater improvement over time in the EORTC Global Health Status in the ChemoRT arm versus the Chemo arm, after adjusting for MSKCC-RPA class and baseline neurologic function status (p=0.005). There was no significant difference in the longitudinal change in any individual scale between the arms (appendix 1E). Disease progression and salvage therapy were associated with longitudinally worse EORTC Global Health status, EORTC Physical Functioning, and EORTC Role Functioning (p<0.001), and higher BN20 Motor dysfunction (p=0.02).

Discussion

In this multicenter randomized study, we investigated the efficacy and toxicity of R-MPV-A with and without LD-WBRT in patients with newly-diagnosed PCNSL. The study demonstrated the addition of consolidation LD-WBRT to R-MPV-A substantially improves PFS in newly-diagnosed PCNSL without increasing cognitive deficits. Results therefore suggest that the markedly reduced dose of WBRT still retains sufficient biological activity to improve disease control in PCNSL.

This study utilized R-MPV-A as the induction chemotherapy, which was found to be highly efficacious and remarkably well tolerated despite the multicenter nature and randomized setting of the trial. Response rates were high (CR of 76% [Chemo] and 100% [ChemoRT]), with excellent PFS and OS observed in both arms, as compared to historical controls^17–19. The reduction in disease burden afforded by the optimized induction regimen may have been crucial to the observed efficacy of LD-WBRT, as consolidation therapies are most effective in the setting of minimal residual disease²⁰. It is also noteworthy that the benefits of LD-WBRT appeared more pronounced in the elderly population, which typically experiences a high rate of relapse and that, unlike younger patients, cannot be easily salvaged, as suggested by the PFS and OS curves in our study (Figure 4)^21–25. However, our study was not designed to investigate this subgroup of patients, nor to detect statistically significant differences in OS, which remains confounded by potentially efficacious salvage treatments. Nevertheless, it is worth noting that the HR for death in the ChemoRT arm was an impressive 0.71.

The development of radiotherapy-related neurotoxicity is an important consideration in treatments involving WBRT^10,26–28. Our trial included neurocognitive testing and quality of life evaluations throughout follow-up. Importantly, such evaluations continued beyond disease progression. This allows for capturing the cognitive deterioration from disease relapse and salvage treatments, in addition to the neurotoxicity related to first-line therapy. The competing risk analysis showed no evidence of a negative impact of LD-WBRT on neurocognitive function, in comparison to chemotherapy alone. Instead, patients in the LD-WBRT arm tended to show better neurocognitive outcomes, despite their more advanced age. Such differences may reflect additional brain tissue injury from more frequent relapses in the Chemo arm, as well as salvage treatments that have included RT and HDCASCT, and that seem more detrimental than the possibly still present radiotherapy-related neurotoxicity. Regardless of arm, tumor progression and salvage therapy were associated with worse neurocognitive outcomes and quality of life, even in successfully salvaged patients. Our analyses therefore underscore the importance of achieving optimal disease control with first-line therapy, and support PFS as an optimal endpoint for studies in newly-diagnosed PCNSL¹⁴.

Other trials have investigated the role of consolidation WBRT in PCNSL, but results are difficult to compare given differing populations and induction regimens. In a study conducted in Germany, 551 patients were randomized to receive MTX +/− ifosfamide with or without WBRT, given at a dose of 45 Gy in 1.5 Gy fractions²⁹. That study found no statistically significant differences in PFS or OS, although outcome was relatively poor in both arms, with median PFS of 18.3 (WBRT arm) and 11.9 (chemotherapy arm) months. Although neuropsychological testing was not performed, clinically defined neurotoxicity was observed in 49% of patients in the WBRT arm and 26% in the chemotherapy arm; no competing risk analysis was performed. Other randomized phase II studies conducted in Europe have sought to compare standard-dose WBRT with HDCASCT, and in general have found that standard WBRT doses are more neurotoxic than HDCASCT, and potentially less efficacious.^30–32 Comparisons to our study are limited because those trials enrolled patients who are candidates for HDCASCT, who tend to be younger, with lower comorbidities and better performance status.

The RTOG has also investigated hyperfractionation as a strategy to reduce neurotoxicity from WBRT³³. In that single-arm study, WBRT was given at a dose of 36 Gy, with 1.2 Gy treatments given twice-daily, in combination with rituximab, methotrexate and temozolomide. The 2-year OS and PFS were 80.8% and 63.6%, respectively. Comprehensive neurocognitive evaluations were not performed but the results showed stable or improved Mini-Mental State Exam scores.

Finally, a retrospective study³⁴ investigated our previously reported LD-WBRT regimen¹¹ in patients younger than 60. Although the LD-WBRT was restricted to patients in CR, results seem in line with those observed in our chemoRT arm, with a 5-year PFS of 69% and OS of 86%. A comparison with patients treated with WBRT at a dose of 40 Gy in 2 Gy fractions showed no differences in efficacy. A subgroup of patients underwent neuropsychological evaluations, which showed no declines in cognitive function. Interestingly, all but one recurrence occurred outside the initially involved site, arguing against the need for focal radiotherapy boosting in this disease.

Our study is the first randomized trial in PCNSL completed in the US, but requires cautious interpretation given its limitations. This includes the phase II trial design and the inherent limitations in sample size, a choice driven by the relative rarity of this disease. A total of 5 patients (11%) in the ChemoRT arm did not receive LD-WBRT (Figure 2), reflecting patterns of care at the time of study design, dedicated treatments for ocular involvement with radiotherapy or intra-ocular chemo or immunotherapy were not utilized upfront. Stratification was performed utilizing the MSKCC-RPA class and not age per se, which resulted in older patients enrolled in the ChemoRT arm (median age of 66 vs 60 in the Chemo arm, p=0.31). Although we achieved high rates of consent to participate in the neurocognitive evaluations (80%), a total 20% of patients refused to participate; those patients were still evaluated for development of neurotoxicity by the treating physicians through standard neurologic exams. Overall, our cognitive evaluations allow for comparisons between the two arms, but is inherently limited in capturing subtle neurotoxic effects of radiotherapy in itself, given the confounding factors of brain injury by disease, relapses and salvage treatments. Finally, further studies are needed to determine how results would compare to other available consolidation options³⁵ such as HDCASCT^30,31,36,37. It must be mentioned that in a single institution setting, R-MPV followed by HDCASCT with thiotepa, busulfan and cyclophosphamide (TBC)³⁸ achieved superior results in comparison to our study, with durable responses and 15-year PFS of 75%.³⁹ Interestingly, a recent randomized phase II study (Alliance 51101), conducted in US institutions similar to ours, investigated HDCASCT with thiotepa and carmustine, versus a non-myeloablative consolidation regimen with etoposide and cytarabine, following an induction regimen of rituximab, methotrexate and temozolomide.¹⁹ With the limitations of cross-study comparisons and the fact that study was terminated earlier due to poor accrual, it appears both arms in our trial achieved comparable or, considering our less favorable population, possibly superior results- for a much more favorable toxicity profile (Appendix 1F). This may reflect the activity and safety of the R-MPV induction, which has consistently achieved excellent results regardless of consolidation choice across multiple other studies^11,21,38.

In summary, our study describes the activity and toxicity of a novel radiotherapy approach to improve disease control in PCNSL without increasing cognitive dysfunction, that could be further explored in elderly or other patients who are not candidates for HDCASCT, or as part of salvage therapies for refractory/ relapsed disease or as a bridge treatment for patients exploring CAR-T cell therapies⁴⁰. Perhaps as importantly, our control arm provides data to guide the treatment for those patients who do not accept any risk of radiotherapy-related neurotoxicity or transplant mortality, and validates R-MPV-A as one of the safest and most efficacious methotrexate-based chemotherapy regimens for PCNSL. In summary, in this phase 2 randomized trial, the addition of LDWBRT meaningfully prolonged PFS, the primary endpoint, and also impacted all of the secondary endpoints in either a statistically or clinically meaningful manner, improving ORR, improving CR, resulting in a lower hazard of death (improving OS), and without cognitive decrement or increased toxicities.

Key Points.

• R-MPV-A was a highly efficacious and safe treatment for PCNSL with or without low-dose radiotherapy.

• Adding low-dose whole brain radiotherapy to R-MPV-A improves progression-free survival.

• Neuropsychological testing showed comparable results, with some tests favoring the chemo-radiotherapy arm.

Importance of the study.

The optimal treatment for PCNSL remains to be established. Previous single-arm studies utilizing induction R-MPV chemoimmunotherapy have achieved excellent results, but the use of consolidation treatments such as radiotherapy or high-dose chemotherapy and stem-cell transplant render their interpretation difficult. One such uncontrolled study utilized substantially reduced doses of whole-brain radiotherapy (LD-WBRT), which was designed to improve local control without excess cognitive decline, and the preliminary dasta provided the hypothesis and data for the presently reported phase 2 randomized study utilizing R-MPV-A chemotherapy with and without LD-WBRT. The study concluded: 1) LD-WBRT contributes significantly to disease control (greater than 2.5 years gain in median PFS), providing a new tool that could be used for consolidation, salvage or bridge treatments 2) Although not statistically significant due to a low event rate, an impressive reduction in the hazard of death (HR = 0.71) 3) Cognitive outcome with LD-WBRT was similar or superior, possibly reflecting more frequent relapses and associated cognitive deterioration with chemotherapy alone, confirming progression-free survival as a key endpoint in this disease and 4) In comparison to historical controls, favorable results were observed in both arms, validating R-MPV-A as one of the most efficacious and safe methotrexate-based chemotherapy regimens for PCNSL, and strongly endorsing the value of LD-WBRT.