In this study, we focused on the influence of age at diagnosis on the prognosis of PCNSL, and age at diagnosis was simply divided into 3 groups to comprehensively analyze the different ages. The median age at the time of diagnosis was 64 years with most patients being older than 60 years (63%) with a trend observed demonstrating higher prevalence in the elderly population (Fig. 1A) also reported in other studies.15 16 17 As in many cancers, age is known as the most widely accepted prognostic marker as well as performance status in PCNSL.18 however in recent large data studies in a few cancers and advanced biological treatments older patients show better response to treatments and survive longer.3, 4 Age cutoffs are controversial with 50 years used as a prognostic score developed by researchers at Memorial Sloan-Kettering Cancer Center19 (MSKCC) and 60 years cutoff used in the International Extranodal Lymphoma Study Group (IELSG) score.12 A recent publication using an alternative age cutoff of 80 stated to enhance the prognostic power compared to other mentioned scales.20 In our cohort of patients older than 60 years, close to 75% died during the study period, however, none of the age groups statistically reached clinically meaningful significance per association with mortality. We did show that delay in diagnosis or relapsed patients had a higher chance of death. The time from symptom appearance to clinical diagnosis was two months in our study and was shown to be a valid prognostic biomarker before.12 Our observation suggests shortening the pretreatment evaluation timeframe, and not preventing treatment in older patients might improve survival.
Delay in diagnosis might stem from nonspecific clinical with a wide array of neurological manifestations shown in Table 1, most commonly progressive cognitive symptoms, and ataxia.
Commonly reported neurological symptoms at presentation such as severe headache, nausea, and vomiting 9, 11 were not as common in our cohort and only occurred in one-third of our patients. Comparison between IVL and PCNSL groups showed the time to histological diagnosis was longer in the IVL group with a median time of 192 days (range: 172–285) versus a median of 38.5 days (range 6-573) p = 0.01. CSF protein levels were not different between. We estimate the prolonged time to diagnosis in IVL was related to the rarity of IVL and to the stroke-like episodes preventing future investigations after the initial stage. Recent publications suggest that random skin biopsy in patients without lymphadenopathy, splenomegaly or bone marrow abnormalities can improve diagnosis time.21
Incidence of PCNSL appears to be increasing with 0.5:1,000,000 cases per year 22, 23 as we have also shown in our cohort, with almost twice as many new cases in 2023 compared to previous years. Incidence in rising mostly in older individuals in parallel to evolving treatments in the past 3 decades. Real-world and population-based studies are still lacking to allow a better understanding of how to predict outcomes and stratify patients clinically and for experimental trials. The wide range of survival for PCNSL patients in our cohort (months to years) emphasizes the need for valid prognostic markers and improved clinical suspicion.
In our study, the cutoff age of 60 as previously published did not reach statistical significance however a threshold of 63 as the age at diagnosis maximally predicted mortality. We show that patients with relapses had higher mortality and were 4 times at risk for death (HR = 4.1), supporting the need to improve the efficacy and durability of responses to 1st line therapy. We have also looked at various prognostic markers for survival as well as risk for relapses. We show that those who presented with cognitive complaints commonly demonstrated involvement of deep brain structures and lower survival rates (p = 0.03), with higher relapse risk (p = 0.04), and. This may be the result of the double involvement of deep brain structures (ie, periventricular regions, basal ganglia, brainstem, and/or cerebellum) as well as the meningeal spread of the brain which are a strategic site for tumor dissemination and were already shown to be significant markers for prognosis.12
We investigated the diagnostic utility for CSF testing as a factor that contributes to delay in diagnosis. CSF studies are routine tests in the evaluation of chronic or subacute neurological presentation such as in cognitive complaints.24 In our cohort the most common CSF abnormality was elevated protein in 80% of cases, with a median CSF protein level of 62 mg/dL. Lymphocytic pleocytosis was present in half of the patients, however, the median number of cells was low (cells/mm). Cytological analysis sensitivity was very low. Other studies showed any CSF abnormality in 80% of PCNSL cases 25 with CSF flow cell type analysis reported to have variable sensitivity of 2–16%. 26 In a review of 3 large studies including more than 150 PCNSL patients with leptomeningeal involvement CSF cytological pathology was noted in only 16% of the patients. 27–29 In our cohort for 93% of patients, the CSF did not help in the diagnostic workup, not even when repeated in 16% of patients. The very low frequency of detection by cytological testing in this study is not self-evident, as seems to be low given the common periventricular location and high rate of leptomeningeal involvement. One likely explanation for this apparent contradiction may be the fragility of the tumor cells in CSF in the changing environment after lumbar puncture, and by prior use of corticosteroids. 30 The ratio of false-negative results may be reduced by increased volume of the CSF specimen to be analyzed (to 10.5 ml) and by immediate processing of the sample.
The overall prognosis was poor with median survival of 3.7 years similar to what has been reported in the literature.5 In our cohort, after relapse median survival was 2 years which is higher than reported. In our cohort, 26 out of 44 patients (59%) experienced relapses in the PCNSL group, while none in the IVL group. Death was reported in 68% by the end of the study. We investigated various factors cited to be relevant for survival12 with neither increased CSF protein level, blood LDH none of which correlated with survival. Shorted time to diagnosis did correlate with reduced survival. Rituximab has been reported to improve outcomes31 however we did not observe this effect in our study like other phase III study by Bromberg et al.32 Currently High-dose methotrexate (HD-MTX) based chemotherapy is considered the cornerstone of PCNSL treatment. In our study Induction treatment with HD-MTX was given to 42 patients (95%), mostly as combination chemotherapy. We found several significant prognostic factors including age, time to diagnosis and relapsed patients which were 4 times more likely to die.
Our work has several limitations, mainly due to the inherent biases of a retrospective study. There was some missing data and loss of follow-up, but not exceeding 10% for most items. Although important, notably in elderly patients, data on toxicity, especially quality of life questionnaires, are lacking. A longer follow-up will be necessary to describe the population of long-term survivors and late relapses.
In summary, PCNSL poses a diagnostic challenge due to its subacute presentation and non-localizing signs, often resulting in diagnostic delay which impacts survival as it prolongs the pretreatment phase. Diagnosis is challenging and the most sensitive diagnostic clues are abnormal brain MRI and brain biopsy supporting diagnosis. CSF had a low sensitivity in the detection of abnormal cells and usually does not lead to diagnosis. Survival seems not directly affected by age at symptom onset; however, by time to treatment initiation with the overall prognosis is poor. Given the rarity of PCNSL long-term follow-up studies are crucial. Understanding the long-term outcomes will guide the development of optimal surveillance strategies and supportive care interventions. Although recent advances have improved our understanding of PCNSL, the need for additional collaborative research is critical.