Characteristics and outcomes of elderly patients with primary central nervous system lymphoma

The Memorial Sloan-Kettering Cancer Center experience




Approximately 50% of all patients with primary central nervous system lymphoma (PCNSL) are aged ≥65 years; however, this group is relatively understudied, and to the authors's knowledge, optimal treatment for older patients is not well defined.


This was a retrospective review of PCNSL patients aged ≥65 years who were treated at Memorial Sloan-Kettering Cancer Center between 1986 and 2008. A multivariate analysis of demographic and clinical variables on prognosis and receipt of treatment was performed.


One hundred seventy-four patients between the ages of 65 and 89 years were identified; there was a slight predominance of women (52.9%). One hundred forty-eight patients were treated with chemotherapy at the time of diagnosis (98% with methotrexate-based therapy) and 31 of these patients also received whole-brain radiotherapy (WBRT). Sixteen patients received WBRT alone. A radiographic response to chemotherapy was noted in 76% of patients. Ninety patients developed disease progression after initial treatment; 74 received salvage therapy and 48% of these patients responded to salvage treatment. The median overall survival was 25 months (range, 18-33 months), and the 3-year survival rate was 36%. Approximately 20.1% of patients were alive for ≥11 years. WBRT was delivered more frequently before 1998, and patients with a history of prior malignancy were less likely to receive WBRT. Age and performance status were identified as the most important predictors of survival. Treatment-related neurotoxicity at 2 years was strongly associated with receipt of WBRT (P = .0002).


PCNSL in the elderly remains sensitive to methotrexate-based chemotherapy and aggressive treatment may be warranted both at the time of diagnosis and disease recurrence. Cancer 2010. © 2010 American Cancer Society.

Primary central nervous system lymphoma (PCNSL) is a rare form of extranodal non-Hodgkin lymphoma (NHL) confined to the brain, cerebrospinal fluid (CSF), spinal cord, leptomeninges, or eyes. PCNSL accounts for <5% of NHLs and approximately 3% of all primary brain tumors.1 Although rare, the age-adjusted incidence in the United States has increased substantially over the past several decades.2 Much of this increase was attributed initially to younger patients with human immunodeficiency virus/acquired immune deficiency syndrome; however, the incidence in this specific population has been decreasing due to the development of active antiretroviral drugs.3 Despite this, the incidence remains high among immunocompetent older patients.4 Moreover, patients aged >60 years account for approximately half of all those diagnosed with PCNSL, making the elderly a priority in the development of improved therapeutic strategies.

In untreated patients the prognosis is poor, with a median survival of 3 to 4 months.5 Whole-brain radiotherapy (WBRT) alone usually results in a median survival of approximately 12 months, and in patients aged >60 years, the median is only 7.6 months.6 Multiple prospective studies using chemotherapy in combination with WBRT have shown improved median survivals of 30 to 60 months.7-15 However, prolonged survival was accompanied by a high rate of neurotoxicity, particularly in the elderly.16 Recent trials have suggested that older patients treated with methotrexate-based chemotherapy and deferred WBRT can achieve high response rates and comparable survival without neurotoxicity, but a shorter progression-free survival.8, 14, 17, 18

Despite these advances, prognosis remains worse in the elderly than in younger patients, even when the same treatment regimen is used in both populations. Age and performance status have been identified as the 2 most important prognostic factors, with age ≥50 years conveying a worse prognosis.19 Age is the strongest prognostic factor, and in some series exerts a stronger effect on survival than treatment regimen.20 In addition, therapeutic advances have not been translated into widespread community practice, particularly among older patients. In a prior study using data from the Surveillance, Epidemiology, and End Results (SEER) cancer registry linked with Medicare claims, we found that only 80% of patients aged ≥65 years with newly diagnosed PCNSL received any treatment.21 Only 43% of treated patients received any chemotherapy, whereas 36% received WBRT alone. Furthermore, advancing age was associated with less therapy and increasing use of WBRT alone, a suboptimal therapeutic approach.18

Balancing treatment efficacy with toxicity is particularly challenging in elderly patients with PCNSL. In this study, we attempt to characterize the prognostic factors of treatment and outcome of a large cohort of older patients treated at Memorial Sloan-Kettering Cancer Center (MSKCC).


Study Design

The MSKCC institutional database was used to identify patients with PCNSL diagnosed between January 1, 1986, and December 31, 2008. Inclusion criteria included histologically confirmed PCNSL and age ≥65 years at the time of diagnosis. The date of the surgical procedure leading to a definitive pathologic diagnosis was used as the date of diagnosis. Data were collected by chart review. We examined the impact of demographic variables on the probability of receiving chemotherapy, WBRT, or both. Response to treatment was defined as complete disappearance of all contrast-enhancing lesions on magnetic resonance imaging or computed tomography when the patient was no longer receiving glucocorticoids (complete response [CR]) or a ≥50% decrease in enhancing disease on stable or decreasing doses of glucocorticoids (partial response [PR]). This retrospective study was approved by the MSKCC Institutional Review Board.

Statistical Analyses

Demographic and clinical variables studied included patient sex, age, Karnofsky performance score (KPS), year of diagnosis, marital status, CSF total protein, deep brain involvement, serum lactate dehydrogenase (LDH) level, hemiparesis, and history of prior cancer. Age at diagnosis was analyzed as a categorical variable. Age trends were assessed using an ordinal age at diagnosis variable. The following variables were dichotomized in the analysis: CSF total protein, KPS, serum LDH, and year of diagnosis. The year 1998 was used as the cutpoint for year of diagnosis because we had eliminated the routine use of WBRT in the initial treatment of older patients by that time. Therapy included any chemotherapy treatment, any radiation treatment, and a combined variable of having been treated with both chemotherapy and RT. Unadjusted logistic regression models were used to analyze demographic and clinical predictors of receipt of treatment. Multivariate logistic regression models were used to analyze those demographic and clinical predictors of receipt of treatment that were univariately significant at the α <.10 level. The association between clinically defined neurotoxicity and WBRT was examined in a landmark analysis at 2 years from diagnosis. Neurotoxicity events occurring after 2 years, as well as patients who were not alive and followed for ≥2 years, were not included in the analysis. The selection of 2 years was based on the median overall survival of the cohort. A landmark analysis was performed to ensure that those included in the analysis had sufficient follow-up to develop neurotoxicity without death as a competing event.

Overall survival was examined from the date of diagnosis to date of death or last follow-up. Univariate Cox proportional hazards models were fit to analyze predictors of survival. A multivariate Cox proportional hazard model was used to analyze predictors of survival that were univariately significant at the α <.10 level. All statistical analyses were performed using SAS statistical software (version 9.1; SAS Institute Inc, Cary, NC).


Patient Characteristics

One hundred seventy-four patients aged ≥65 years who were diagnosed with PCNSL between 1986 and 2008 were included (Table 1). The median age at diagnosis was 72 years (range, 65-89 years); there was a predominance of women (52.9%). Most patients were diagnosed before 1998 (62.6%), and a majority of patients were married (67.2%). Forty-six (26.4%) patients had a history of a prior nonlymphomatous malignancy.

Table 1. Patient Demographics
CharacteristicNo. of Patients%
  1. KPS indicates Karnofsky performance status; CSF, cerebrospinal fluid; LDH, lactate dehydrogenase.

Age at diagnosis, y  
Y of diagnosis  
 Prior to 199810962.6
 1998 and after6537.4
Marital status  
KPS at presentation  
CSF total protein, mg/dL  
Deep brain involvement  
Serum LDH, U/L  
History of prior malignancy  

Treatment at Diagnosis

One hundred seventeen patients (67.2%) received chemotherapy alone, 16 (9.2%) patients received RT alone, and 31 (17.8%) patients received both chemotherapy and RT (Table 2). Six patients received no treatment, and data regarding therapy were missing for 4 patients. Of the 148 patients who received chemotherapy, 145 (98%) received high-dose methotrexate (at a dose of 3.5 g/m2); response to chemotherapy (CR + PR) was noted in 76% of these patients.

Table 2. Initial Treatment for Elderly Patients With PCNSL
  1. PCNSL indicates primary central nervous system lymphoma.

Chemotherapy alone11767.2
Radiotherapy alone169.2
Chemotherapy + radiotherapy3117.8
No treatment63.5

Predictors of Treatment

Significant univariate predictors of receipt of chemotherapy were age at diagnosis, year of diagnosis, and deep brain involvement. Patients who were diagnosed in 1998 and later were more likely to receive chemotherapy (odds ratio [OR], 3.4; 95% confidence interval [95% CI], 1.3-8.7) as were patients with deep brain involvement (OR, 4.6; 95% CI, 1.2-17.3). There was also a trend noted with age in which older patients were more likely to receive chemotherapy (P = .07). In the multivariate analysis for predictors of chemotherapy treatment, only deep brain involvement was found to be significant (OR, 4.5; 95% CI, 1.1-17.7), adjusting for age at diagnosis and year of diagnosis.

Significant univariate predictors of receipt of RT included age at diagnosis, year of diagnosis, marital status, deep brain involvement, and history of prior cancer. Patients were significantly less likely to receive RT if they were older, diagnosed in 1998 or later, unmarried, had deep brain involvement, or had at least 1 prior cancer. In the multivariate setting, those patients who were diagnosed in 1998 or later were less likely to receive RT (hazard ratio [HR], 0.2; 95% CI, 0.1-0.5) as were those with at least 1 prior cancer (HR, 0.2; 95% CI, 0.1-0.9). Marital status remained of borderline significance (P = .06) as did a trend for age (P = .06). Other results for the analyses of predictors of PCNSL treatment can be seen in Table 3.

Table 3. Predictors of Receiving Preradiation Chemotherapy, Radiotherapy, and Both Modalities for the Elderly PCNSL Cohort
VariableLevelChemotherapyRadiotherapyBoth Radiotherapy and Chemotherapy
Unadjusted OR (95% CI)PaAdjustedb OR (95% CI)PaUnadjusted OR (95% CI)PaAdjustedb OR (95% CI)PaUnadjusted OR (95% CI)PaAdjustedb OR (95% CI)Pa
  • PCNSL indicates primary central nervous system lymphoma; OR, odds ratio; 95% CI, 95% confidence interval; KPS, Karnofsky performance status; CSF, cerebrospinal fluid; LDH, lactate dehydrogenase.

  • a

    P value is for trend for age variable.

  • b

    Adjusted for other variables in the multivariate logistic model. Variables were entered into the multivariate model if univariately significant (at α <0.1 level).

  • c

    No estimate because there were no subjects who received both preradiation chemotherapy and radiotherapy who were aged ≥80 years.

Age categories, y65-691.
 70-742.2 (0.7-6.6) 3.4 (0.7-17.2) 0.7 (0.3-1.6) 0.9 (0.3-2.6) 1.0 (0.4-2.4) 1.7 (0.6-5.2) 
 75-792.4 (0.7-8.0) 1.7 (0.4-7.4) 0.4 (0.2-1.0) 0.5 (0.2-1.7) 0.5 (0.2-1.6) 0.9 (0.3-3.0) 
 ≥803.9 (0.5-32.5) 2.4 (0.3-21.4) 0.1 (0.0-0.8) 0.2 (0.0-1.4) No estimatec No estimatec 
SexMale1.0.31  1.0.69  1.0.30  
 Female0.6 (0.2-1.6)   0.9 (0.4-1.7)   0.7 (0.3-1.5)   
Y of diagnosisPrior to 19981.0<.011.0.991.0<.00011.0.00051.0<.00011.0.0001
 1998 and after3.4 (1.3-8.7) 1.1 (0.3-3.3) 0.1 (0.0-0.2) 0.2 (0.1-0.5) 0.1 (0.0-0.3) 0.1 (0.0-0.3) 
Marital statusMarried1.0.43
 Unmarried0.7 (0.3-1.7)   0.4 (0.2-1.0) 0.4 (0.1-1.0) 0.2 (0.1-0.7) 0.2 (0.1-0.8) 
KPS<701.0.14  1.0.46  1.0.81  
 ≥700.4 (0.1-1.3)   1.3 (0.6-2.7)   1.1 (0.5-2.5)   
CSF total protein, mg/dL≤601.0.11  1.0.85  1.0.68  
 >600.2 (0.0-1.5)   1.1 (0.4-3.1)   0.8 (0.3-2.5)   
Deep brain involvementNo1.  
 Yes4.6 (1.2-17.3) 4.3 (1.1-17.2) 0.4 (0.2-1.0) 0.8 (0.3-2.0) 0.5 (0.2-1.3)   
Serum LDH, U/L≤2001.0.29  1.0.75  1.0.72  
 >2003.4 (0.3-34.5)   1.2 (0.4-3.3)   1.2 (0.4-3.6)   
HemiparesisNo1.0.46  1.0.28  1.0.91  
 Yes0.7 (0.3-1.8)   1.5 (0.7-3.1)   0.9 (0.4-2.3)   
History of prior cancerNo1.0.54
 Yes0.7 (0.2-2.1)   0.3 (0.1-1.0) 0.2 (0.1-1.0) 0.2 (0.1-1.1) 0.2 (0.0-1.0) 

Survival Analysis

The median overall survival for the entire cohort was 25 months (95% CI, 18-33 months) (Fig. 1). The 3-year survival rate was 36% (95% CI, 29%-44%). Ninety patients (52%) developed disease progression with a median time to disease progression of 24 months (95% CI, 15-30 months); 74 of these 90 patients (82.2%) received salvage therapy, 53 received chemotherapy alone, 14 received RT, and 7 received both. Radiographic response (CR + PR) was noted in 48% of patients receiving salvage treatment. Thirty-nine patients were alive at a median follow-up of 34 months, and their treatment was comparable to the group as a whole. Thirty-five (89.7%) of these patients were diagnosed before 1998 and represent long-term survivors; these 35 patients were 20.1% of the entire cohort. Age at diagnosis and KPS were found to be the only 2 significant predictors of overall survival in the univariate analysis (Table 4). In the multivariate model, increasing age demonstrated a significant trend (P = .03) toward a greater probability of death. Patients with a KPS ≥70 were significantly less likely to die (P <.0001).

Figure 1.

(Top) Overall survival of the Memorial Sloan-Kettering Cancer Center cohort of 174 older patients with primary central nervous system lymphoma is shown. (Bottom) Overall survival from the Surveillance, Epidemiology, and End Results (SEER)-Medicare study is shown. DOD indicates died of disease; 95% CI, 95% confidence interval. Reproduced with permission from Panageas KS, Elkin EB, Ben-Porat L, DeAngelis LM, Abrey LE. Patterns of treatment in older adults with primary central nervous system lymphoma. Cancer. Vol. 110, No. 6, 2007, 1338-1344.21 ©2007 American Cancer Society. This material is reproduced with permission of Wiley-Liss, Inc, a subsidiary of John Wiley & Sons, Inc.

Table 4. Predictors of Overall Survival for the Elderly PCNSL Cohort Treated at MSKCC
VariableLevelUnadjusted HR (95% CI)PaAdjustedb HR (95% CI)Pa
  • PCNSL indicates primary central nervous system lymphoma; MSKCC, Memorial Sloan-Kettering Cancer Center; HR, hazard ratio; 95% CI, 95% confidence interval; CSF, cerebrospinal fluid; KPS, Karnofsky performance status; LDH, lactate dehydrogenase.

  • a

    P value was for trend for age variable.

  • b

    Adjusted for other variables in the multivariate logistic model. Variables were entered into the multivariate model if univariately significant (at α <0.1 level).

 Female1.0 (0.7-1.4)   
Age categories, y65-691.
 70-741.1 (0.7-1.7) 1.1 (0.7-1.7) 
 75-791.3 (0.8-2.0) 1.3 (0.8-2.1) 
 ≥802.6 (1.4-4.8) 2.0 (1.1-3.7) 
Y of diagnosisPrior to 19981.0.77  
 1998 and after0.9 (0.7-1.3)   
Marital statusMarried1.0.40  
 Unmarried1.2 (0.8-1.7)   
CSF total protein, mg/dL≤601.0.29  
 >601.3 (0.8-2.0)   
Deep brain involvementNo1.0.44  
 Yes1.2 (0.8-1.7)   
 ≥700.5 (0.3-0.7) 0.5 (0.3-0.7) 
Serum LDH, U/L≤2001.0.72  
 >2001.1 (0.7-1.7)   
 Yes1.2 (0.8-1.7)   
History of prior cancerNo1.0.80  
 Yes0.9 (0.6-1.5)   


Acute treatment-related toxicity was not collected prospectively in all patients. However, NCI CTCAEv 3.0 grades 3 and 4 nephrotoxicity related to high-dose methotrexate was noted in 13 of the 145 patients treated (9%); this is a minimal incidence. There was no grade 5 nephrotoxicity.

There were 77 patients alive 2 years from diagnosis. Eighteen (23%) developed neurotoxicity, 67% of whom received RT. This is in contrast to 55 patients who survived 2 years and never developed neurotoxicity, 18% of whom received RT (P = .0002). There were 4 patients alive at 2 years who developed neurotoxicity after that timepoint and whose neurotoxicity events were excluded.


The results of the current study demonstrate that vigorous therapy of elderly patients with PCNSL can produce prolonged survival in a significant percentage. The data from the current study represent a selected patient population that may not reflect the elderly population in the community; however, nearly 10% of our patients were aged ≥80 years and nearly 40% had a KPS of ≤60 at the time of diagnosis. Despite these clinical features, 96.5% of our patients received therapy, which is a much higher percentage than noted in the SEER-Medicare cohort or the general British Columbia population, most likely reflecting our vigorous approach to these patients.21, 22 Furthermore, 83% (145 of 174 patients) were treated with high-dose methotrexate. In general, treatment was well tolerated and the majority responded. Even at the time of disease recurrence, 48% of patients responded to salvage therapy. These data demonstrate that many elderly patients can, and should, receive a treatment regimen comparable to that which is often reserved for younger patients.

Age is the most influential prognostic factor in PCNSL and invariably affects treatment decisions.19, 21, 23 The SEER-Medicare data indicated that 20% of all older patients are not treated at all.21 In a large population-based study in British Columbia, 10.7% of all PCNSL patients seen in that province between 1990 and 2003 received steroids only and no definitive treatment; the majority were older patients.22 Furthermore, when treatment was administered, it was often suboptimal. WBRT was the sole therapy for 36% of patients in the SEER-Medicare database, 17% received chemotherapy alone, and 26% received both.21 The median survival of the SEER-Medicare cohort was 7 months, comparable to the 7.6 months achieved with WBRT alone in patients aged ≥60 years.6, 21 Outcome is superior when chemotherapy is part of the initial treatment plan. In a comparably aged population, high-dose methotrexate-based chemotherapy with or without WBRT achieved a median overall survival of 29 months in those patients aged >60 years, which is similar to the 25-month median survival noted in our cohort.18

The majority of patients in our elderly cohort received chemotherapy alone as initial treatment. Approximately 83% of our total population and 98% of those who received any chemotherapy received methotrexate (at a dose of 3.5 g/m2). This reflects the aggressive approach of a tertiary referral center accustomed to using high doses of chemotherapy for this disease despite the poor performance status of many patients. For many older patients, it has been well demonstrated that methotrexate-based regimens are tolerable and effective.24-27 One study demonstrated no significant difference between methotrexate (at a dose of 4 g/m2)-related toxicities in patients aged ≥60 years versus those aged <60 years.26 Similarly, a recent retrospective review of patients aged ≥70 years who were treated with high-dose methotrexate (at a dose of 8 g/m2) demonstrated that this regimen was well tolerated and produced radiographic response rates of >90%, and that few patients required discontinuation of chemotherapy for toxicity.27 Although numerous studies have demonstrated the efficacy of methotrexate-based regimens, to the best of our knowledge no phase 3 trial has been conducted to date to demonstrate the superiority of any particular regimen. However, methotrexate appears to be well tolerated in the elderly, supporting its use in these patients.

The late complications of combined modality therapy, namely neurotoxicity and dementia, were first recognized in the late 1990s. The results of the current study demonstrated significantly less use of WBRT in patients treated in 1998 or later. Despite this trend, the year of treatment appeared to have no effect on survival, suggesting that chemotherapy alone may be a good choice for the initial treatment of PCNSL in the elderly. Radiographic responses, time to disease progression, and overall survival were found to be similar to the results of prior studies using methotrexate-based regimens.12, 17, 18, 24, 25, 27 Recent data suggest that consolidation with low doses of RT after immunochemotherapy may improve upon survival without the risk of neurotoxicity.28, 29 However, longer follow-up and additional studies are needed.

Other than year of diagnosis, few sociodemographic characteristics were found to predict treatment. Married patients were more likely to receive both chemotherapy and RT, which was also found in the SEER-Medicare dataset, perhaps suggesting that patients with strong social support are more likely to receive aggressive therapy. Our sample size was relatively small and may not have been powered to detect small but significant associations between other patient characteristics and treatment received.

The current study has several limitations, including the inherent biases of a retrospective study. Other important variables including treatment toxicity, quality of life measures, and detailed neurocognitive assessments on all patients were not available. Referral and treatment practice biases are almost certainly inherent in this single-center cohort. Nonetheless, this study reflects a relatively large group of older patients with a rare disease, specifically examining patterns of care, prognostic factors, and outcomes of specialized treatment practices as they apply to the elderly PCNSL population.

Although outcome is worse in older patients with PCNSL, that should not preclude the use of aggressive therapy because the majority of patients respond and improve clinically. The patients in the current study cohort had a favorable time to disease progression, and greater than half responded to salvage therapy. Clinical trials aimed at this population are crucial to improving outcomes associated with this disease, and patients should be referred to institutions with an interest in PCNSL for inclusion into these studies.


The authors made no disclosures.