Demographic information and treatments
This study included a total of 40 consecutive treatment-naïve patients with PCNSL. All patients had B-cell lymphoma and 39 patients were diagnosed by brain biopsy or vitrectomy. One patient was confirmed to have a mature B-cell origin by FCM.
Ocular involvement was present in 15 (37.5%) patients, six (15.0%) of whom had exclusive ocular involvement at presentation. The remaining nine (22.5%) patients had synchronous ocular and brain involvement. Only one patient with leptomeningeal subtype was diagnosed by FCM in this study.
The median age was 58 (range, 24–79) years and 18 (45%) patients were male. Twenty-two (55%) and nine (22.5%) patients were categorized as intermediate- and high-risk according to International Extranodal Lymphoma Study Group score (IELSG)[18].
Six patients with PVRL received rituximab plus lenalidomide (R2 regimen) as first-line treatment while the remaining 34 patients received high-dose methotrexate and rituximab-based regimens. Further treatment details are described in Table 1.
The overall response rate (ORR) of the whole series was 80.0% (32/40), and the complete response (CR) rate was 52.5% (21/40). The median follow-up period was 13.6 (range:2–55) months. At the last follow-up, 24 (60%) patients had relapsed and 12 had died. The median PFS and OS were 617.7 and 1,146 days, respectively (Fig. 1)
CSF IL-10 concentration at pretreatment, interim-induction and end-induction
The first lumbar puncture was performed before treatment in all patients. Eighteen patients had increased intracranial pressure (>180 mmH2) and protein levels were increased in 34/40 patients. Atypical cells were found in 7/40 patients by light microscopy.
The median IL-10 concentration of all patients before treatment was 31.7 pg/mL (range: 5.0–1000) and 34 samples had detectable IL-10 levels (>5.0. pg/mL). According to our previous report[12], we evaluated the results in this larger cohort and found specificities of 85% (34/40) and 90% (36/40), for a CSF IL-10 concentration cutoff of >8.2 pg/mL or a CSF IL-10/IL-6 ratio >0.72, respectively.
The mean IL-10 concentration was higher in patients with CSF involvement (560.2 ± 174.9 vs. 136.5 ± 41.3 pg/mL, p = 0.0113). We observed a higher trend in the multiple-lesion group compared to that in the single-lesion group (275.5 ± 68.42 vs. 59.23 ± 37.96 pg/mL, p = 0.053) (Fig. 2) The involved sites (PVRL vs. brain involvement, deep-lesion group vs. superficail-lesion group) and IELSG score (low-, intermediate-, or high-risk) were not related to IL-10 concentration.
The concentrations of CSF IL-10 after two cycles of treatment were tested in 38/40 patients, except for two patients who refused additional LP. The IL-10 concentrations decreased in all but four patients. The mean IL-10 concentration decreased significantly after treatment (220.9 ± 53.62 vs. 67.49 ± 29.1 pg/mL, 95%CI 41.04–265.8, p = 0.0088) with a decline of 153.4 pg/mL by paired t-test. The four patients with increased CSF IL-10 levels all relapsed within 6 months (range: 1–6 months).
We observed that a higher IL-10 level after 2 cycles was associated with poor PFS (300 vs. 507 days, p 0.0703), but not with OS (p = 0.553).
The IL-10 concentrations at the end of induction were measured in 29/40 patients (Fig. 3). Eight patients did not finish induction therapy due to early relapse or severe adverse reactions, and three patients refused LP.
At the end of induction therapy, 21 patients achieved a complete remission (CR), 3 achieved partial remission (PR), 1 patient was stable (SD), and 4 patients had progressive disease (PD). The mean IL-10 concentrations of the CR, PR, and non-response (SD+PD) subgroups were 47.91 ± 37.22, 286.3 ± 120.9, and 219.3 ± 195.2 pg/mL, respectively. There was a trend of lower IL-10 concentration in the CR group compared to that in the non-CR group (47.91 ± 37.22 vs. 244.4 ± 123.8 pg/mL, p = 0.0504).
An issue of concern was the relationship between sustained CSF IL-10 concentration and ocular involvement. Six of 21 patients who achieved CR had detectable CSF IL-10 concentrations; all had eye involvement at the time of diagnosis and five had the PVRL subtype. Although the baseline IL-10 concentration was lower in the patients with PVRL, the end-induction of IL-10 concentration was higher in PVRL than that for CNS-involved patients who achieved CR (28.84 ± 17.37 vs. 5.0 pg/mL, p = 0.0645).
CSF IL-10 concentrations increased before disease relapse
Among 24 patients who experienced relapse during the follow-up, 19 had undergone CSF IL-10 measurement. The sensitivity of detectable CSF IL-10 was 78.95% (15/19) in relapsed patients and the mean concentration was lower than the pretreatment level (ΔIL-10: 139.1 ± 295.6 pg/mL, p = 0.055, paired t-test).
16 patients underwent dynamic CSF surveillance during induction and follow-up. CSF analysis was performed at each induction cycle, which took 3 or 4 weeks, and every 3 months during the follow-up. MRI surveillance was performed every two cycles during the induction phase and every 3 months during the follow-up. Once the patients developed any symptoms that suggested disease recurrence, CSF analyses and MRI scans were performed immediately. We defined IL-10 relapse as a positive conversion in patients with previously undetectable IL-10 and increased concentration compared to the last test in patients with sustained IL-10 levels. As described previously, disease relapse was defined as new CNS lesions on MRI scan and/or new intra-ocular lesions by ophthalmologic examination. IL-10 relapse was detected a median of 67 (range: 28–402) days earlier than disease relapse in 10 patients, while disease relapse occurred at the same time in three patients. Three patients relapsed without detectable IL-10 levels (Fig. 4, Table 2).
Prognostic factors for survival
Cox univariate regression was applied for PFS prognosis analysis using clinical characteristics including ocular involvement, PVRL subtype, IELSG score, treatment regimen, CSF IL-10 levels after two cycles, and post-treatment CSF IL-10 levels. Only detectable CSF IL-10 levels after induction therapy were poor predictors of PFS (p = 0.049, HR 0.638, 95% CI 0.408–0.998). The same factors were evaluated by univariate analysis using COX for OS. The high-risk group of IELSG showed a trend of poor OS (p = 0.077, HR 0.343, 95%CI 0.104–1.124).