SEARCH

SEARCH BY CITATION

Keywords:

  • primary central nervous system lymphoma;
  • incidence;
  • survival;
  • treatment

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

BACKGROUND.

The incidence of primary central nervous system lymphoma (PCNSL) appears to be increasing in some countries, whereas it is stable in others. Many reports the last decades have suggested that there have been improvements in the treatment of PCNSL. The objective of this study was to analyze time trends in the incidence, clinical features, histologic diagnosis, treatment, and outcome of nonacquired immunodeficiency syndrome (non-AIDS) PCNSL in Norway from 1989 to 2003.

METHODS.

Patients were identified by a chart review of all patients who had a recorded diagnosis of PCNSL from 1989 to 2003 in The Norwegian Cancer Registry. The histologic and cytologic material from each patient was re-examined by pathologists. Time trends were analyzed according to year of diagnosis grouped into 3 5-year periods: 1989–1993, 1994–1998, and 1999–2003.

RESULTS.

There were 98 patients who had confirmed, newly diagnosed non-AIDS PCNSL in Norway from 1989 to 2003. The incidence rate increased during the consecutive 5-year periods from 0.89 per million during 1989 to 1993, to 1.74 per million during 1994 to 1998, and to 1.82 per million during 1999 to 2003 (P = .013). Diagnostic delay and overall survival did not improve with time. Survival decreased from 1999 to 2003 compared with survival from 1994 to 1998, which was explained in part by reduced performance status and fewer patients receiving combined chemotherapy and radiotherapy during 1999 to 2003. In multivariate analysis, age ≤50 years, a good performance status, and active treatment (especially combined chemotherapy and radiotherapy) significantly improved survival.

CONCLUSIONS.

The incidence of PCNSL is increasing in Norway. Despite diagnostic and therapeutic advances over the last decades, neither a reduction in diagnostic delay nor any improvement in overall survival with time was observed. The search for improved understanding of etiology and treatment should be intensified. Cancer 2007. © 2007 American Cancer Society.

The incidence of primary central nervous system lymphoma (PCNSL) seems to increase in some geographic regions but not in others. It reportedly has increased in the United States,1–3 the United Kingdom,4 the Netherlands,5 and Japan6 but has remained stable in Canada,7 Denmark,8 Scotland,9 Hong Kong,10 and India.11, 12 In a previous study from Norway, we observed a nonsignificant trend toward increased incidence from 1989 to 1998 (P = .069).13

The treatment of PCNSL has changed considerably the last decades. Throughout the 1980s, whole-brain radiotherapy was the mainstay of treatment.14 This normally induced temporary tumor regression with a reported median survival of 12 to 18 months15, 16 and a 5-year survival rate <20%.16 In later reports, combined chemotherapy and radiotherapy resulted in a median survival of 33 to 60 months and 5-year survival rates as high as 22% to 40%.13, 17–19 The disadvantage of combined-modality therapy is the high risk of late neurotoxicity, which affects from 26% to 58% of patients and especially affects those aged >60 years.19–22 Surgical resection yields no survival benefit and may increase functional deficits.23, 24

PCNSL may be difficult to diagnose. Common debut symptoms, such as personality changes and headache,23 are unspecific, and findings on diagnostic imaging may be variable or inconclusive.25–27 In a previous study, we observed that there was a considerable delay in the diagnosis of PCNSL.28 However, considering the advances in diagnostic imaging technologies the last decades and the reported improvements in treatment for PCNSL, we hypothesized that there has been a reduction in diagnostic delay and that overall survival has improved for patients with PCNSL over time in Norway.

In the current retrospective national survey, we investigated time trends in incidence, clinical features, histologic diagnosis, treatments, outcomes, and identified prognostic factors for all patients who had histologically verified, nonacquiredimmunodeficiency syndrome (non-AIDS) PCNSL diagnosed in Norway from 1989 to 2003. Few comparable population-based time-trend analyses of PCNSL have been published14, 29; and, to our knowledge, this is the first study to be based on unselected and complete national material.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

The regional committee for research ethics approved this study. In Norway (population, 4.2 million in 1989, 4.6 million in 2003), there is mandatory cancer reporting to the Norwegian Cancer Registry, which makes cancer registration especially comprehensive. A list of all patients who were diagnosed with PCNSL in Norway from January 1, 1989 to December 31, 2003 was obtained from this national registry (n = 180 patients). To review the complete medical records of the patients, the first author (I.S.H.) visited the 6 major university hospitals, which diagnosed and treated all but 8 of these patients. Medical records from other treating hospitals were sent to the authors by mail. Clinical features, histological diagnosis, radiologic findings, medical treatments, and outcomes were registered, and pathologic material was reviewed. Patients who had histologically verified, non-AIDS PCNSL after a chart review and pathologic review were included (Fig. 1).

thumbnail image

Figure 1. Inclusion of patients after chart review according to 5-year periods. PCNSL indicates primary central nervous system lymphoma; HIV, human immunodeficiency virus.

Download figure to PowerPoint

All available and relevant histologic and cytologic material from all non-AIDS patients who had conclusive or possible PCNSL as their primary histologic diagnosis (n = 104 patients) was reviewed by 1 pathologist (B.K.K.) and, for patients who were treated at Rikshospitalet University Hospital (n = 38 patients), also by another pathologist (D.S.). The material was reclassified according to the 2001 World Health Organization (WHO) classification.30 Supplementary immunohistochemical and histochemical staining techniques (Giemsa stain, Periodic acid-Schiff, and reticulin stain) were performed when required to confirm the diagnosis and to reclassify the lymphomas. At a minimum, the following antibodies were used for immunohistochemistry: CD20, CD79a, CD3, and Ki-67. The following markers were used as required: CD45 (leukocyte common antigen), CD43, CD30, CD138, CD23, CD10 (common acute lymphoblastic leukemia antigen), CD5, cyclin D1, CD8, CD4, CD68, CD34, terminal deoxynucleotidyl transferase, CD56 (neural cell adhesion molecule), K light chain, and Λ light chain.

In total, we included 98 non-AIDS patients who had a final diagnosis of PCNSL based on review of biopsy and/or autopsy material (n = 96 patients) or on cytology (n = 2 patients; material was available for review in only 1 patient). Three of these patients who were diagnosed between 1994 and 1998 and 2 patients who were diagnosed between 1999 and 2003 received immunosuppressive medication for other medical conditions (ulcerative colitis, renal transplantation, rheumatoid arthritis, and Wegener granulomatosis) before the first symptoms that were attributable to PCNSL. Because of the pathologic review and updated records from the Norwegian Cancer Registry and the treating hospitals, the current sample has undergone minor changes from our previous studies of PCNSL diagnosed in Norway between 1989 and 1998.13, 28

For patients who were diagnosed while they were alive (n = 80 patients), performance status at the time of histologic diagnosis was specified in 39 patients and was estimated retrospectively from the medical records in 41 patients. The cause of death was established based on medical records and death certificates (available in 81 of 89 patients who died).

Statistical Analyses

We calculated incidence rates based on the number of patients with newly diagnosed PCNSL each year and the population of Norway at January 1 of the same year. Trend tests and 95% confidence intervals (95% CI) for incidence were calculated according to the Poisson distribution. To estimate the average change in yearly incidence rates, the estimated annual percent change (EAPC) was calculated by fitting a least-squares regression line to predict the natural logarithm of the yearly rates.

The Mantel-Haenszel exact test was applied to analyze time trends in dichotomous variables concerning clinical features and treatment regimens for the different 5-year periods. A nonparametric linear trend test (the Jonckheere-Terpstra test) was applied to analyze time trends in continuous variables (age at the time of diagnosis, time from initial symptom to histologic diagnosis, and whole-brain and total radiation doses). The time of histologic diagnosis was the date of diagnostic procedure or the date of death in patients who were diagnosed at autopsy.

The duration of survival was calculated as the time between the date of diagnostic procedure (operation, biopsy, and/or cytology) that led to histologic diagnosis and either the date of death or January 1, 2007. Survival curves were calculated and plotted using the method of Kaplan and Meier. Log-rank tests were used for univariate comparison of survival between groups. The Cox proportional-hazards model was used to study the effect on survival of several variables simultaneously and to estimate hazard ratios. The proportional-hazard assumption was assessed by comparing the Cox model with a time-dependent model. All P values in the analyses are 2-sided. The analyses were performed with Stata and SPSS software.

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

Incidence, Diagnosis, and Treatment in the Total Sample

The average annual incidence rate of non-AIDS PCNSL in Norway from 1989 to 2003 (n = 98 patients) was 1.49 per million (95% CI, 1.21–1.82 per million). The median duration of symptoms before histologic diagnosis was 13 weeks (range, 1–209 weeks). The median patient age at the time of diagnosis was 68 years (range, 11–83 years). The diagnosis was established first at autopsy in 18 of 98 patients (18%). Before death, imaging indicated brain tumor in 13 of 18 patients (72%) and no tumor in 5 of 18 patients (28%); a diagnostic biopsy was inconclusive at the time in 3 patients and was not performed in the remaining patients. Only 1 patient received radiotherapy (30 grays [Gy]), and 3 patients received chemotherapy (cytarabine in 1 patient, cyclophosphamide in 1 patient, and combined alkylating agents and vincristine in 1 patient).

Primary treatments received by the 80 patients who were diagnosed while they were alive included only chemotherapy in 20 patients (25%), only radiotherapy in 14 patients (18%), combined chemotherapy and radiotherapy in 30 patients (38%), and only symptomatic therapy in 16 patients (20%)—all but 1 of those 16 patients had a poor performance status (WHO 2–4) at the time of diagnosis. Most chemotherapy regimens (66%) included high-dose methotrexate (HD MTX) (>1000 mg/m2) either alone (28%) or combined with other chemotherapeutic drugs (38%). Radiotherapy included whole-brain irradiation in all 44 patients (100%) and focal boost toward macroscopic tumor in 17 of 44 patients (39%). Additional chemotherapy and additional radiotherapy were received by 6 patients and 4 patients, respectively, at the time they developed clinical recurrence from 4 to 36 months (median, 18 months) after primary treatment.

Outcome in Total Sample

Of 64 patients who received active therapy, 43 patients (67%) initially tolerated the treatment well. In 21 patients (33%), the treatment was complicated by deep venous thrombosis (n = 4 patients), septicemia (n = 4 patients), nephrotoxicity (n = 3 patients), pneumonia (n = 3 patients), meningitis (n = 2 patients), pulmonary embolism (n = 1 patient), pulmonary edema (n = 1 patient), intracerebral hemorrhage (n = 1 patient), agranulocytosis (n = 1 patient), or aplastic anemia (n = 1 patient). In 7 patients, these complications, which included septicemia (n = 2 patients), nephrotoxicity (n = 1 patient), pneumonia (n = 2 patients), pulmonary embolism (n = 1 patient), and pulmonary edema (n = 1 patient), were fatal. Six of those deaths were related to chemotherapeutic treatment, including only chemotherapy (n = 5 deaths) and combined chemotherapy and radiotherapy (n = 1 death), and 1 death was related to steroid medication.

Eighty-nine of 98 (91%) patients died. The cause of death was progression of PCNSL in 71 patients and complications related to the initial treatment in 7 patients. Four patients (including 1 patient who had had a concomitant intracranial recurrence) died of systemic lymphoma from 3 to 8 years after initial therapy. None of those patients had signs of systemic lymphoma at the initial staging with bone marrow biopsy, x-ray of the chest and abdominal ultrasound/computed tomography (CT) studies. Six patients died with late neurotoxicity from 1 to 10 years after primary treatment without signs of progressive PCNSL. In 1 patient, the cause of death was unknown.

For patients who were diagnosed while they remained alive (n = 80 patients), the overall estimated survival from the time of diagnosis was 46% at 1 year, 38% at 2 years, and 16% at 5 years; and the median survival was 7 months (Fig. 2). Among patients who were diagnosed at death (n = 18 patients), the median survival from the time of initial symptoms was 3 months, and the 1-year survival rate was only 6%.

thumbnail image

Figure 2. Survival curve with 95% confidence interval for patients who were diagnosed with primary central nervous system lymphoma while they were alive from 1989 to 2003 (n = 80).

Download figure to PowerPoint

Time Trends in Incidence, Clinical Features, and Histologic Diagnosis

The incidence increased from 0.89 per million during 1989 to 1993, to 1.74 per million during 1994 to 1998, and to 1.82 per million during 1999 to 2003 (P = .013). The incidence increased the first part of the study period, was relatively stable between 1993 and 2002, and had a peak in 2003 (Fig. 3). The annual incidence rate also was increasing (P = .002) with an EAPC of 7.3%.

thumbnail image

Figure 3. The number of patients with newly diagnosed primary central nervous system lymphoma in Norway from 1989 to 2003.

Download figure to PowerPoint

The number of patients who had their histologic diagnosis established while they were alive in the consecutive 5-year periods (63%, 87%, and 85%, respectively) tended to increase (P = .09) (Table 1). The WHO performance status tended to decline in the last 5-year period (P = .09). Other patient characteristics did not change significantly. In particular, the time to diagnosis was not reduced during the study period (P = .32) (Table 1).

Table 1. Patient Characteristics in the 3 5-Year Periods (n = 98)
 1989–1993, n1 = 191994–1998, n2 = 381999–2003, n3 = 41P*
No. of patients%No. of patients%No. of patients%
  • CT indicates computed tomography; MR, magnetic resonance; WHO PS, World Health Organization performance status.

  • *

    Mantel-Haenszel exact test for linear association.

  • linear trend test: Jonckheere-Terpstra test.

Categorical variables       
Sex      .79
 Women126317452254 
 Men73721551946 
Symptoms prior to diagnosis
 Focal neurologic126318472049.42
 Personality change105323612868.27
 Cerebellar symptoms/vertigo84217451229.27
 Headache842174518441.00
 Visual disturbance632381639.22
 Epilepsy21151337.66
Lesions on CT/MR studies      .46
 No detectable expansive lesion2112537 
 Single136821552151 
 Multiple42115401742 
Patients diagnosed while alive126333873585.09
Morphologic diagnosis established while alive by n1 = 12, n2 = 33, n3 = 35      .78
 Open biopsy86720612366 
 Stereotactic or endoscopic biopsy32512361131 
 Cytology of cerebrospinal fluid181313 
Autopsy performed after death among patients diagnosed while alive, n1 = 11, n2 = 28, n3 = 29      .62
 Yes32727414 
 No87326932586 
WHO PS at histologic diagnosis among patients diagnosed while alive, n1 = 12, n2 = 33, n3 = 35      .09
 0–13251855514 
 2–497515463086 
Continuous variables
Age at diagnosis, y
 Mean66.8 63.1 64.5  
 Median (range)69.9 (32–81) 69.6 (11–83) 65.5 (15–83) .51
Time to diagnosis, wk
 Mean40.1 17.4 21.9  
 Median (range)11.9 (2–209) 11.8 (1–76) 15.1 (2–84) .32

There was an increasing use of CT/ultrasound (chest/abdomen) studies, bone marrow biopsy, cerebrospinal fluid examination, and ophthalmologic examination at the initial diagnostic work-up (Table 2). X-rays of the chest were obtained frequently in all periods. No significant changes in histologic characteristics were observed over the different 5-year periods (Table 3).

Table 2. Examinations Undertaken at Initial Diagnostic Work-up in the 3 5-Year Periods (n = 98)
Examinations undertakenNo. of patients (%)P*
1989–1993, n1 = 191994–1998, n2 = 381999–2003, n3 = 41
  • CT indicates computed tomography.

  • *

    Mantel-Haenszel exact test for linear association.

X-ray of the chest18 (95)37 (97)39 (95)1.00
CT of the chest2 (11)18 (47)34 (83)<.001
Abdominal ultrasound and/or CT of the abdomen and pelvis12 (63)32 (84)41 (100)<.001
Bone marrow biopsy8 (42)28 (74)31 (76).03
Cerebrospinal fluid examination7 (37)20 (53)27 (66).04
Ophthalmologic examination6 (32)8 (21)23 (56).02
Table 3. Revised Histologic Diagnosis According to the 2001 World Health Organization Classification
Morphologic characteristicsNo. of patients (%)P*
1989–1993, n1 = 191994–1998, n2 = 381999–2003, n3 = 41
  • PTLD indicates posttransplantation lymphoproliferative disease; NOS, not otherwise specified.

  • *

    Mantel-Haenszel exact test for linear association.

  • Histologic or cytologic material was not available for review in this patient.

Large B-cell lymphomas
 Diffuse14 (74)29 (76)34 (83).42
 Monomorphic PTLD-diffuse large B-cell0 (0)1 (3)0 (0)1.00
 Intravascular0 (0)0 (0)1 (2).61
B-cell lymphoma, NOS3 (16)3 (8)3 (7).36
Precursor B-lymphoblastic lymphoma0 (0)0 (0)1 (2).61
Burkitt lymphoma/Burkitt-like lymphoma0 (0)1 (3)2 (5).46
Peripheral T-cell lymphoma1 (5)2 (5)0 (0).25
Malignant lymphoma, probable1 (5)2 (5)0 (0).25

Time Trends in Primary Treatment and Outcome

HD MTX monotherapy was more common in the 2 first periods, and HD cytarabine and alkylating agents were used more often in the latter periods (Table 4). A focal radiation boost was given infrequently in the latter period. The average total dose of radiation decreased significantly with mean doses of 45.6 Gy, 43.1 Gy, and 33.6 Gy, respectively, in the 3 consecutive 5-year periods (P = .003).

Table 4. Treatment and Causes of Death Among Patients Diagnosed While Alive in the 3 5-year Periods (n = 80)
 No. of patients (%)P*
1989–1993, n1 = 121994–1998, n2 = 331999–2003, n3 = 35
  • HD indicates high dose; MTX, methotrexate; CHOP, combined cyclophosphamide, doxorubicin, vincristine, and prednisone; Gy, grays; PCNSL, primary central nervous system lymphoma.

  • *

    P values are from Mantel-Haenszel exact test for linear association unless indicated otherwise.

  • Nonparametric linear trend tests (Jonckheere-Terpstra tests).

  • One fatal primary treatment-related complication.

  • §

    Two fatal primary treatment-related complications.

  • HD cytarabine: mean (median; range) dose, 3.1 g (2 g; 1–8 g) ×2; mean (median; range) number of cycles, 2 (2; 1–4).

  • Three fatal primary treatment-related complications.

  • #

    Alkylating agents: carmustine (n = 7), lomustine (n = 3), procarbazine (n = 7), trophospamide (n = 1).

  • **

    Etoposide (n = 2), idarubicin (n = 2), thiotepa (n = 1).

  • ††

    MTX in 96% (27 of 28 patients): mean absolute dose (median; range): 12.9 mg (12 mg; 12–15 mg); mean number of cycles (median; range), 3.8 (4; 1–12). Cytarabine in 25% (7 of 28 patients): mean dose (median; range), 44.3 mg (50 mg; 30–50 mg); mean number of cycles (median; range), 3.3 (3; 1–6).

  • ‡‡

    Cyclophosphamide (n = 3), doxorubicin (n = 3), vincristine (n = 3), prednisone (n = 2), carboplatin (n = 2), etoposide (n = 1), temozolomide (n = 1).

  • §§

    Mean (median; range) total dose, 35.4 Gy (40 Gy; 20–41.4 Gy; n = 4).

  • ||

    Including fatal primary treatment-related initial complications.

Primary treatment
Only symptomatic therapy4 (33)5 (15)7 (20).56
Only chemotherapy1 (8)8 (24)11 (31).15
Only radiotherapy4 (33)3 (9)7 (20).69
Combined chemotherapy and radiotherapy3 (25)17 (52)10 (29).63
Chemotherapy: n1 = 4, n2 = 25, n3 = 21
 Intravenous therapy4 (100)24 (96)20 (95)1.00
 HD MTX2 (50)17 (68)14 (67)§.81
  HD MTX monotherapy2 (50)11 (44)1 (5).005
  Absolute dose and cycle: n1 = 2, n2 = 17, n3 = 14
  Mean absolute dose; median [range], g1.8; 1.8 [1.7–2]5.3; 5.0 [3.4–12]6.6; 6.4 [1.8–12].01
  Mean no. of cycles; median [range]4; 4 [2–6]3.4; 4 [1–6]4.3; 4 [1–12].32
 CHOP2 (50)2 (8)1 (5).06
 HD cytarabine0 (0)6 (24)10 (48).03
 Alkylating agent#0 (0)5 (20)11 (52)§.007
 Vincristine2 (50)9 (36)15 (71)§.07
 Other**0 (0)0 (0)4 (19)§.04
 Intrathecal therapy††1 (25)15 (60)12 (57).65
Radiotherapy: n1 = 7, n2 = 20, n3 = 17
 Whole-brain radiation7 (100)20 (100)17 (100) 
 Spinal radiation0 (0)1 (5)0 (0)1.00
 Booster toward tumor3 (43)13 (65)1 (6).015
 Completed radiotherapy6 (86)18 (90)13 (77).57
 Palliative intent (after progression on chemotherapy)0 (0)4 (20)4 (24).28
 Total mean dose; median [range], Gy45.6; 46.0 [39–50]43.1; 50.0 [20–50]33.6; 39.6 [3–45].003
 Mean whole brain dose; median [range], Gy41.8; 40.0 [39–50]36.5; 40.0 [20–41]33.0; 39.6 [3–45].07
Primary treatment-related toxicity; n1 = 8, n2 = 28, n3 = 28
 Treatment-related initial complications4 (50)7 (25)10 (36).85
 Fatal primary treatment-related initial complications1 (13)2 (7)4 (14).78
 Late neurotoxicity2 (25)4 (14)4 (14).62
Additional treatment at clinical recurrence: n1 = 12, n2 = 33, n3 = 35
 Chemotherapy‡‡1 (8)3 (9)2 (6).77
 Radiotherapy§§0 (0)1 (3)3 (9).29
Causes of death: n1 = 12, n2 = 33, n3 = 35
 PCNSL||9 (75)22 (67)29 (83).37
 Systemic lymphoma0 (0)3 (9)1 (3)1.0
 Delayed neurotoxicity2 (17)4 (12)0 (0).04
 Unknown0 (0)0 (0)1 (3).59
 Alive1 (8)4 (12)4 (11)1.0

Survival did not improve with time; rather, survival was significantly lower from 1999 to 2003 compared with survival from 1994 to 1998 (P = .015). The middle period (1994–1998) showed a trend toward better survival compared with the other 2 5-year periods (P = .06) (Fig. 4). No patients (0 of 35 patients) who were diagnosed with PCNSL from 1999 to 2003 died of delayed neurotoxicity compared with 2 of 12 patients (17%) and 4 of 33 patients (12%) who were diagnosed from 1989 to 1993 and from 1994 to 1998, respectively (P = .04) (Table 4).

thumbnail image

Figure 4. Survivals curve for patients who were diagnosed while alive from 1989 to 1993 (12 patients, 1 censored), from 1994 to 1998 (33 patients, 4 censored), and from 1999 to 2003 (35 patients, 4 censored) (log-rank test, P = .06).

Download figure to PowerPoint

Prognostic Factors in Patients Diagnosed While Alive

Primary treatment, age at diagnosis, WHO performance status, and HD MTX-containing regimens were highly significant prognostic factors (P < .006), and the year of diagnosis and total radiation dose were almost significant prognostic factors (P = .06 and P = .08, respectively) in the univariate analyses (Table 5). For patients who received only symptomatic treatment (n = 16 patients), only chemotherapy (n = 20 patients), only radiotherapy (n = 14 patients), and combined chemotherapy and radiotherapy (n = 30 patients), the estimated median survival was 0.7 months (95% CI, 0.5–1.0 months), 3.7 months (95% CI, 0.0–7.3 months), 8.8 months (95% CI, 0.0–19.0 months), and 43.5 months (95% CI, 22.1–64.8 months), respectively (Fig. 5).

thumbnail image

Figure 5. Survival curves according to treatment modalities for patients who were diagnosed while alive from 1989 to 2003 (n = 80). Only symptomatic treatment, 16 patients (0 censored); only chemotherapy, 20 patients (3 censored); only radiotherapy, 14 patients (0 censored); combined chemotherapy and radiotherapy, 30 patients (6 censored) (log-rank test, P < .001).

Download figure to PowerPoint

Table 5. Univariate Analysis Results for Patients Diagnosed While Alive From 1989 to 2003
VariableNo. of patientsMedian survival (95% CI), moP*
  • 95% CI indicates 95% confidence interval; WHO PS, World Health Organization performance status; CT, computed tomography; MR, magnetic resonance; HD MTX, high-dose methotrexate; Gy, grays; CSF, cerebrospinal fluid.

  • *

    Log-rank test.

Primary treatment, n = 80  <.001
 No treatment (only symptomatic)160.7 (0.5–1.0)
 Only chemotherapy203.7 (0.0–7.3)
 Only radiotherapy148.8 (0.0–19.0)
 Chemotherapy and radiotherapy3043.5 (22.1–64.8)
Age at diagnosis, y; n = 80  <.001
 ≤5011≥96
 >50695.2 (2.1–8.4)
WHO PS, n = 80  <.001
 0–12649.2 (29.1–69.3)
 2–4543.0 (1.1–4.8)
Sex, n = 80  .55
 Men359.4 (0.0–23.1)
 Women456.4 (0.0–13.1)
Year of diagnosis, n = 80  .06
 1989–1993125.4 (0.0–13.2)
 1994–19983331.9 (0.0–68.4)
 1999–2003354.1 (0.0–8.9)
Time to diagnosis, wk; n = 80  .77
 <12405.4 (2.7–8.1)
 ≥12408.8 (0.0–23.7)
Operation prior to diagnosis, n = 77  .89
 Tumor resection458.8 (0.0–23.8)
 Biopsy327.0 (0.0–21.4)
Lesions on CT/MR studies, n = 76  .20
 Single498.8 (0.0–24.2)
 Multiple277.0 (0.0–15.2)
HD MTX-containing regimens, n = 50  .006
 Yes3338.6 (18.5–58.7)
 No175.2 (0.6–9.9)
HD cytarabine-containing regimens, n = 50  .63
 Yes167.1 (0.0–41.0)
 No3429.7 (7.6–51.9)
Alkylating agent-containing regimens, n = 50  .11
 Yes167.0 (0.0–30.2)
 No3434.0 (13.3–54.6)
Vincristine-containing regimens, n = 50  .82
 Yes2623.1 (1.9–44.3)
 No2423.8 (0.0–61.7)
Total radiation dose in primary treatment, Gy; n = 44  .08
 <452516.7 (0.0–40.1)
 ≥451943.5 (20.3–66.7)
Malignant cells in CSF, n = 41  .48
 No3621.4 (11.4–31.4)
 Yes55.2 (0.0–10.9)
Raised protein level in CSF, n = 33  .99
 No73.0 (0.0–7.1)
 Yes2615.1 (0.0–30.9)

A multivariate Cox proportional-hazards model that included the variables age (>50 years vs ≤50 years), WHO performance status (2–4 vs 0–1), and treatment (chemotherapy, radiotherapy, or combined chemotherapy and radiotherapy vs no treatment) (n = 80 patients) indicated that all variables had a significant independent impact on survival (Table 6). When the year of diagnosis was included in the model, this variable was not significant (P = .33).

Table 6. Multivariate Cox Proportional Hazards Model for Patients Diagnosed While Alive From 1989 to 2003 (n = 80)
VariablePHR (95% CI)
  1. HR indicates hazard ratio; 95% CI, 95% confidence interval; WHO PS, World Health Organization performance status.

Age (>50 y vs ≤50 y)<.0019.8 (3.10–31.21)
WHO PS, 2–4 vs 0–1.0471.8 (1.01–3.36)
Primary treatment<.001 
 No treatment, only symptomatic 1.00
 Chemotherapy vs no treatment 0.21 (0.10–0.46)
 Radiotherapy vs no treatment 0.11 (0.05–0.26)
 Combined chemotherapy and radiotherapy vs no treatment 0.05 (0.02–0.12)

When radiotherapy was included as part of primary treatment (no vs yes), HD MTX-containing regimen (no vs yes), age (>50 years vs ≤50 years), and WHO performance status (2–4 vs 0–1) in a model that was restricted to patients who received chemotherapy (n = 50 patients), radiotherapy significantly improved survival (hazard ratio, 4.05; 95% CI, 1.92–8.53; P < .001); whereas HD MTX did not reach statistical significance (hazard ratio, 1.50; 95% CI, 0.76–2.96; P = .24). In a similar model that was restricted to patients who received radiotherapy (n = 44 patients), radiation dose (<45 Gy vs ≥45 Gy) was not significant (hazard ratio, 1.33; 95% CI, 0.53–3.36; P = .55).

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

In the current 15-year survey, patients from a national registry had their diagnosis of PCNSL validated by a thorough review of the medical records and pathologic material. This led to the exclusion of 82 of 180 patients (46%) because of human immunodeficiency virus-related PCNSL (n = 23 patients), incorrect registration of PCNSL (n = 41 patients), or lack of histologic verification (n = 18 patients). Exclusion rates >50% after review of medical records in other studies7, 31 further underscore the importance of validating data from registries. Because our survey was retrospective, it did not influence the practice or outcomes we wished to assess. Only 9 patients (9%; 2 patients in 1994–1998 and 7 in 1999–2003) had participated in a prospective clinical trial32; thus, the time trends we present are almost unaffected by any study and reflect ordinary clinical practice in Norway.

We observed significantly increasing incidence rates of PCNSL in Norway from 0.89 per million during 1989 to 1993 to 1.82 per million during 1999 to 2003 (P = .013). We believe that this estimate is fairly reliable. Registry of cancer is mandatory in Norway, and no changes concerning the registration of PCNSL have been carried out that possibly may have biased our findings. The proportion of excluded patients also has been rather stable (39%–55%) in the different 5-year periods, indicating that there were no obvious changes in registration practice. The observed peak in new diagnoses of PCNSL in 2003 (Fig. 3) was odd but may have been caused by statistical variation within a rising trend. Omitting 2003 data from the analysis, we still observed significant increasing incidence rates during 1989 to 2002 (P = .023). The rising trend also seems to continue, because 2004 had the second greatest number of recorded PCNSL diagnoses in the Norwegian Cancer Registry since 1989 (n = 15 patients in 2004, n = 20 patients in 2003).

Patients who were treated in the last 5-year period went through more adequate staging regimens than patients who were treated in the first 5-year period. Thus, the chance of falsely including patients with occult systemic lymphoma has declined, strengthening our findings of increased incidence. This chance also is balanced by the opposite, small chance of falsely excluding some of the patients who had inconclusive histologic diagnoses (Fig. 1).

It has been suggested that the observed increase in the incidence of PCNSL may have been caused in part by improved technologies available for the diagnosis of brain tumors, including CT and magnetic resonance imaging (MRI).2 However, Olson et al. reported increasing PCNSL rates in the United States that coincided with stable glioma rates when both tumors were being diagnosed with similar imaging technology.1 In Norway, the rates of central nervous system tumors almost doubled from 1970 to 1999.33 However, between 1989 and 2003, the increase in primary brain tumors (excluding tumors in meninges, corpus pineale, hypophysis, and cranial nerves) was moderate.34 The average annual incidence rate was 85.7 per million (95% CI, 83.4–87.9 per million), and the EAPC was 2.3% (95% CI, 1.7–2.9%; P < .001) (Fig. 6). Because the corresponding EAPC of PCNSL in our material (n = 98 patients) was 7.3% (95% CI, 2.6–12.1%; P = .002), PCNSL increased more than primary brain tumors. Although lack of a review of the medical records and histologic material from all patients with primary brain tumors introduces the risk of misclassification that limits our ability to truly compare these EAPCs, it seems unlikely that the impact of improved diagnostic tools would explain the increasing incidence of PCNSL in Norway.

thumbnail image

Figure 6. Incidence rates of primary central nervous system lymphoma (PCNSL) and primary brain tumors in Norway 1989–2003. EAPC indicates estimated annual percent change.

Download figure to PowerPoint

The increased incidence of PCNSL also exceeds what may be caused by an increasing incidence of non-Hodgkin lymphomas in general. The average annual incidence rate of non-Hodgkin lymphoma in Norway during the studied period was 143.9 per million (95% CI, 141.0–146.8 per million)34 with an EAPC of only 1.8% (95% CI, 1.3–2.2%; P < .001)—far less than the EAPC for PCNSL (7.3%).

Because immunocompromised patients with transplants or autoimmune conditions are predisposed to PCNSL,35 the frequent use of immunosuppressive medication in recent years theoretically may account for an increasing incidence of PCNSL. However, according to our material, only 5 patients received immunosuppressive medication before the first symptoms that were attributable to PCNSL. When these patients were omitted from the analyses, there remained a significant increase in the annual incidence rate (P = .006) and in incidence rates between the 5-year periods (P = .021).

The time to diagnosis, surprisingly, was stable over the 3 5-year periods despite the increased numbers of whole-body MRI scanners in Norway (6 scanners, 13 scanners, 28 scanners, and 73 scanners on January 1, 1989, 1994, 1999, and 2004, respectively), allowing the widespread use of MRI as a diagnostic tool in addition to CT. Thus, increased availability and reported advances in diagnostic imaging of PCNSL36–38 did not shorten the diagnostic delay of PCNSL in Norway over the 15-year study period.

Performance status was estimated retrospectively in 41 of 80 patients (51%), and this may have reduced the accuracy of the variable. However, like in many other studies, good performance status and young age had a favorable impact on survival13, 24, 39–42; patients who received only symptomatic therapy had shorter survival18–20 and combined chemotherapy and radiotherapy was superior to chemotherapy alone43 and radiotherapy alone.44 We observed an estimated median survival of 44 months with combined chemotherapy and radiotherapy compared with a median survival of 9 months with radiotherapy alone and 4 months with chemotherapy alone, which led to fatal complications in 5 of 20 patients (25%). These survival rates are in the lower range of those reported by others.15, 16, 18, 19, 41, 42, 44 In part, this may be because there are fewer patients aged ≤50 years in population-based material like that used in the current study compared with institution-based materials (eg, 14% of patients in our material compared with 30% of patients in material from a tertiary referral center42). Although it was not significant in our material, the omission of HD MTX in the chemotherapy regimen tended to reduce survival (hazard ratio, 1.50; 95% CI, 0.76–2.96; P = .24), which is in accordance with studies that reported a significant impact of HD MTX with hazard ratios of 2.329 and 1.48.44

Despite the therapeutic advances reported from clinical trials the last decades,17, 20 the overall survival of patients who were diagnosed with PCNSL while they remained alive over the 3 5-year periods did not improve over time. This is in accordance with 2 population-based studies from the United States (1975–1999)14 and Canada (1990–2003),29 which reported stable outcomes in the periods studied. By contrast, in a large study from Japan (1985–1999),45 the investigators reported improved survival with time.

Survival was reduced significantly from 1999 to 2003 compared with survival from 1994 to 1998 (P = .015). One likely explanation is that fewer patients who were diagnosed while alive had a good performance status during 1999 to 2003 compared with patients who where diagnosed during 1994 to 1998 (14% vs 55%, respectively; Pearson chi-square test: P< .001). In addition, fewer patients received combined chemotherapy and radiotherapy as primary treatment from 1999 to 2003 (29% vs 52%; chi-square test: P = .05). This may have been caused by a growing concern regarding the well-documented high risk of neurotoxicity associated with combined chemotherapy and radiotherapy.17, 19, 21 To minimize this complication, primary treatment with chemotherapy alone has been proposed, with radiotherapy delayed until there is evidence of disease recurrence in complete responders.17, 32 However, the addition of radiotherapy in patients who were receiving chemotherapy improved survival significantly in our patients (hazard ratio, 4.05; 95% CI, 1.92–8.53; P < .001).

Furthermore, the mean total dose of radiotherapy given declined over time from 43.0 Gy during 1994 to 1998 to 33.5 Gy during 1999 to 2003 (P = .002). This may have been caused by reports in the 1990s indicating that dosages >40 Gy to 50 Gy did not increase survival15, 46 but were associated with significant late neurotoxicity,20 which made some authorities recommend lower radiation doses.15 However, reducing the dose (from 45 Gy to 30.6 Gy) in patients aged <60 years may be associated with an increased risk of recurrence and lower overall survival.47 In our material, the univariate analysis indicated a trend toward increased survival with radiation doses ≥45 Gy (P = .08) (Table 5), although this trend was weakened in the multivariate analysis (hazard ratio, 1.33; 95% CI, 0.53–3.36; P = .55).

In summary, the current results demonstrate that there has been an increasing incidence of non-AIDS PCNSL in Norway. Despite diagnostic and therapeutic advances over the last decades, we observed neither a reduction in diagnostic delay nor any improvement in survival with time. Rather, survival during the last 5-year period of the study decreased compared with the preceding 5-year period, which may be explained in part by the worsened performance status and fewer numbers of patients receiving combined chemotherapy and radiotherapy. In the material as a whole, the variables young age (≤50 years), good WHO performance status (0–1), and active treatment had an independent favorable impact on survival. Improved treatment and care must be implemented in ordinary clinical practice if survival rates comparable to those reported from some clinical studies are to be achieved. Considering the increasing incidence and the dismal prognosis of PCNSL in most patients, improved treatment strategies and continued search for etiologic factors are required.

Acknowledgements

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

We thank all institutions that made it possible for us to review their medical records.

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES