Diffuse large B-cell lymphoma (DLBCL) is the most common aggressive subtype of non-Hodgkin lymphoma (NHL) in adults, accounting for 30–40% of cases worldwide (1). Although frontline regimens such as R-CHOP have improved clinical outcomes, approximately 2–5% of patients still develop secondary central nervous system lymphoma (SCNSL) (2), with rates increasing to 10–12% in high-risk subgroups (3). SCNSL is a rare but often fatal complication, typically occurring within 7–8 months of initial diagnosis (4–6). Patients with central nervous system (CNS) involvement exhibit dismal prognoses, with a median overall survival (OS) of only 3.4 months (7). Given the poor outcomes associated with CNS relapse, early identification of high-risk DLBCL patients is crucial for guiding preventive CNS-directed interventions and improving survival.

In 2016, Schmitz et al. (8) developed the Central Nervous System International Prognostic Index (CNS-IPI) as a tool for stratifying CNS relapse risk in DLBCL patients. The CNS-IPI has since been incorporated into National Comprehensive Cancer Network (NCCN) guidelines to assess the likelihood of CNS involvement (9). The reported 2-year CNS relapse rates were 0.6%, 3.4%, and 10.2% for patients in low-, intermediate-, and high-risk groups, respectively (8). However, follow-up data revealed that 55% of patients who developed CNS relapse belonged to the high-risk group, 40% to the intermediate-risk group, and 5% to the low-risk group (8). Moreover, real-world data over a 3-year follow-up period reported a higher-than-expected CNS relapse rate of 10.1% in the intermediate-risk category (5). These findings underscore the limited sensitivity of CNS-IPI in accurately identifying DLBCL patients at elevated risk of SCNSL. To improve predictive accuracy, incorporation of biologically relevant biomarkers has been proposed.

Recent studies have highlighted the role of cytokines within the tumor microenvironment. Interleukin-10 (IL-10), an anti-inflammatory cytokine produced by monocytes, macrophages, and tumor cells, contributes to the regulation of the blood–brain barrier and promotes lymphoma progression, thereby facilitating CNS infiltration (1, 10). Interleukin-6 (IL-6), a key growth factor for B cells, supports tumor growth and metastasis through its secretion by DLBCL cells and the surrounding stroma (11). While the diagnostic value of cerebrospinal fluid (CSF) IL-10 and IL-10/IL-6 ratios has been established in primary CNS lymphoma (PCNSL) (12–15), their utility in predicting CNS relapse in SCNSL remains underexplored. Unlike PCNSL, which presents with isolated CNS involvement, SCNSL requires early risk stratification before CNS dissemination occurs, highlighting the need for minimally invasive predictive biomarkers. Prior studies have shown significantly elevated serum IL-6 and IL-10 levels in DLBCL patients compared to healthy individuals (10, 16), with elevated cytokine levels correlating with shorter progression-free survival (PFS) and OS (10). However, the prognostic and predictive value of serum IL-6 and IL-10—alone or in combination with CNS-IPI—remains insufficiently characterized in the context of SCNSL.

In this study, we analyzed serum and CSF levels of IL-10 and IL-6 at diagnosis to assess their clinical significance in predicting disease progression and CNS relapse in DLBCL. We also examined correlations between serum and CSF cytokine levels and evaluated whether incorporating these biomarkers into the CNS-IPI model could enhance its predictive performance for SCNSL.

SCNSL represents a devastating complication in DLBCL with uniformly poor outcomes. The CNS-IPI does not account for the biological heterogeneity of DLBCL (2), limiting its ability to effectively stratify patients at risk of CNS involvement. Previous reports have linked elevated serum levels of IL-6 and IL-10 to adverse prognosis in DLBCL (25, 26). In the present study, we investigated the clinical utility of serum IL-6, IL-10, and the IL-10/IL-6 ratio, analyzing their associations with disease progression and SCNSL development. Our findings demonstrate that the serum IL-10/IL-6 ratio functions as an independent risk factor for both disease progression and CNS relapse. Moreover, incorporating this biomarker into CNS-IPI and NCCN-IPI models significantly improved their ability to identify high-risk patients.

Chronic inflammation creates a microenvironment conducive to tumor initiation, promotion, progression, metastasis, and resistance to chemotherapy, primarily through the activity of inflammatory cells, cytokines, chemokines, and other mediators. Key cytokines within this milieu include IL-6, IL-10, among others (27). Elevated serum IL-6 levels have been consistently reported in patients with DLBCL compared to healthy controls (28, 29). Dlouhy et al. (30) demonstrated that higher serum IL-6 levels were significantly associated with shorter PFS in DLBCL. More recently, Bao et al. (10) showed that increased IL-6 not only correlates with adverse clinical features and PFS but also functions as an independent prognostic marker for relapse and OS. Compared with patients exhibiting normal serum IL-6 levels, those with elevated IL-6 were more likely to exhibit partial or no response to therapy.

Similarly, several studies have reported that elevated serum IL-10 at diagnosis is associated with unfavorable disease features and poor clinical outcomes (31–33). In comparison with healthy controls, DLBCL patients demonstrated significantly higher serum IL-10 levels (p < 0.001). Patients with IL-10 concentrations ≥5.0 pg/mL experienced significantly shorter PFS and OS (p < 0.001) (10). Yi et al. (1) investigated the association between pretreatment serum cytokines and the occurrence of SCNSL in DLBCL. Their findings revealed that patients with elevated serum IL-10 levels were more likely to develop SCNSL and experienced earlier CNS involvement. Moreover, CNS infections or inflammatory conditions can elevate IL-10 levels in CSF, often accompanied by a reduced IL-10/IL-6 ratio (34, 35). This may reflect a compensatory anti-inflammatory response by immune cells in reaction to pro-inflammatory cytokines such as IL-6 (12, 36). Therefore, in this study, we also evaluated the IL-10/IL-6 ratio in both CSF and serum to determine its utility in identifying patients at high risk for SCNSL.

Our study explored the prognostic significance of serum IL-10 and IL-6 in the development of SCNSL and uncovered several clinically relevant insights. Although both cytokines were assessed, only serum IL-10 levels were significantly elevated in SCNSL patients (30.80 vs. 3.00 pg/mL, p = 0.001), suggesting a predominant role in CNS tropism. Importantly, the serum IL-10/IL-6 ratio emerged as a particularly powerful biomarker, demonstrating an 8.6-fold increase in SCNSL patients (7.80 vs. 0.91 pg/mL, p = 0.001). This ratio may better reflect the immunosuppressive tumor microenvironment than absolute cytokine levels alone, as the anti-inflammatory effects of IL-10 may dominate when disproportionately elevated relative to IL-6. Multivariate analysis confirmed that a serum IL-10/IL-6 ratio ≥2.3 was the only independent predictor of both poor PFS (HR = 7.30, p = 0.010) and SCNSL development (odds ratio [OR] = 43.20, p = 0.001). The strength of these associations highlights the ratio’s potential clinical relevance. Supporting this, patients with elevated ratios had significantly shorter median PFS (15.42 vs. 37.94 months, p = 0.002), representing a clinically meaningful 22.5-month difference. These findings support a novel paradigm in which the prognostic impact of IL-10 in DLBCL is modulated not solely by its absolute level, but by its balance relative to IL-6.

The Central Nervous System International Prognostic Index (CNS-IPI) is currently the most widely used tool for predicting CNS involvement in DLBCL. In our cohort of 56 patients, CNS-IPI stratification—though limited by a median follow-up of 23.5 months (range: 0.7–41.9 months) and sample size—did not demonstrate statistically significant differences in SCNSL incidence among the low-, intermediate-, and high-risk groups (p = 0.136). Similarly, no significant difference in time to CNS relapse was observed between high-risk and non–high-risk groups. To improve risk prediction, we developed a modified scoring system, termed CNS-IPI-ratio, which integrates the serum IL-10/IL-6 ratio with the CNS-IPI score. Risk levels in the original CNS-IPI were assigned 1, 2, and 3 points for low, intermediate, and high risk, respectively; an additional point was assigned for a serum IL-10/IL-6 ratio ≥2.30. Patients scoring 1–2 points were classified as non–high-risk, while those scoring 3–4 points were categorized as high-risk. Our stratification approach yielded two key clinical findings. First, the CNS-IPI-ratio high-risk group identified significantly more SCNSL cases than the non–high-risk group (85.7% vs. 41.9%, p = 0.045), outperforming conventional models in discriminative capacity. Second, high-risk patients developed CNS relapse nearly 18 months earlier than non–high-risk patients (median 22.32 vs. 40.35 months, p = 0.017), indicating that CNS-IPI-ratio may predict both the likelihood and timing of CNS progression. The National Comprehensive Cancer Network International Prognostic Index (NCCN-IPI) remains the standard prognostic model for DLBCL in the rituximab era (22). However, this model relies solely on clinical parameters and excludes biological markers. In our analysis, the NCCN-IPI alone showed a non-significant trend toward PFS and OS differences between high-risk and non–high-risk groups. After incorporating the serum IL-10/IL-6 ratio into the NCCN-IPI, the modified scoring system yielded statistically significant differences in both PFS and OS. These findings suggest that integrating the serum IL-10/IL-6 ratio significantly enhances the prognostic accuracy of the NCCN-IPI.

Tumor RNA-seq analyses across multiple DLBCL cohorts consistently demonstrated that elevated IL-10/IL-6 expression ratios were associated with poorer clinical outcomes. Notably, an integrated IPI-ratio model provided superior prognostic discrimination compared to the conventional IPI alone. Although tumor mRNA expression data from GEO cohorts support the biological relevance of our serum biomarker findings, they serve a complementary rather than substitutive role. These transcriptomic analyses reinforce the mechanistic basis and cross-cohort consistency of IL-10/IL-6 dysregulation, yet the clinical utility of the serum IL-10/IL-6 ratio must be directly validated in our ongoing prospective study, as serum protein levels in an independent cohort remain the gold standard for biomarker evaluation. The consistent demonstration of IL-10/IL-6 imbalance in both the peripheral circulation (serum) and tumor microenvironment (RNA-seq) underscores the biological significance of this cytokine axis in DLBCL pathogenesis. While incorporating the IL-10/IL-6 ratio into existing clinical models enhances risk stratification, these preliminary findings warrant prospective validation in multicenter cohorts before clinical implementation.

The diagnostic value of elevated CSF IL-10 remains debated. Ikeguchi et al. (37) reported that although median CSF IL-10 concentrations were significantly higher in patients with CNS lymphoma (including PCNSL and SCNSL) compared to those with inflammatory demyelinating CNS diseases (p < 0.01), multivariate analysis indicated that CSF IL-10 lacked independent predictive value. In contrast, numerous studies have supported the utility of elevated CSF IL-10 in distinguishing PCNSL from other brain tumors, neuroinflammatory conditions, and systemic DLBCL (38–41). Notably, increased CSF IL-10 has been proposed as a potential biomarker for CNS lymphoma, including SCNSL (14, 15). On the other hand, IL-6 is frequently detectable in CSF during viral, bacterial, and fungal infections, as well as CNS inflammatory diseases—especially bacterial meningitis (42, 43). Patients with CNS inflammatory disorders had significantly higher CSF IL-6 concentrations than those with PCNSL (p = 0.032) (41). Most studies evaluating the CSF IL-10/IL-6 ratio have focused on PCNSL, with this ratio frequently serving as a diagnostic marker. For example, a CSF IL-10/IL-6 ratio >1 was detected in 75% (58/77) of PCNSL cases (44), and a similar threshold in vitreous or atrial fluid strongly suggested primary vitreoretinal lymphoma (PVRL) (45). Additionally, studies by Song (12) and Li et al. (46) reported that the CSF IL-10/IL-6 ratio could aid in distinguishing PCNSL from systemic non-Hodgkin lymphoma. When comparing the CSF IL-6/IL-10 ratio between CNS inflammatory disease and PCNSL, the ratio was significantly higher in the former (p < 0.001) (41). In our study, CSF IL-10/IL-6 ratios were significantly elevated in SCNSL patients relative to both non-SCNSL DLBCL patients and healthy controls. However, given the small number of SCNSL patients with available CSF samples (n = 3), these findings should be interpreted as preliminary and hypothesis-generating rather than definitive. While these observations are biologically plausible and consistent with prior PCNSL research, larger, multicenter prospective studies are needed to establish their diagnostic validity.

Regarding the serum IL-10/IL-6 ratio, our findings identified it as an independent risk factor for both progression-free survival (PFS) and SCNSL development in DLBCL. Serum IL-10 levels have previously been shown to reflect systemic tumor burden (1). Additionally, IL-10 plays a pivotal role in the JAK/STAT signaling pathway and can induce PD-L1 expression, facilitating immune evasion by tumor cells (1, 47). IL-10 is also implicated in blood-brain barrier dysfunction, promoting CNS infiltration by malignant cells (1, 16, 48). In contrast, IL-6—being a pleiotropic cytokine—is elevated in nearly all infectious or inflammatory conditions (42). Therefore, an elevated IL-10/IL-6 serum ratio may help exclude patients with underlying systemic infection or inflammation, thereby improving specificity for tumor-associated cytokine imbalances. A higher IL-10/IL-6 ratio at initial diagnosis may thus serve as a potential biomarker for identifying DLBCL patients at high risk for SCNSL. However, some studies have reported contradictory findings. Zou (49) observed no significant difference in serum IL-10/IL-6 ratios between DLBCL patients with and without SCNSL. Similarly, Geng et al. (38) found no statistically significant difference in this ratio between PCNSL patients and controls (patients with intracranial metastatic tumors or gliomas). These discrepancies may be attributable to differences in sample size, disease stage, or clinical characteristics across cohorts. Given the collective evidence on serum and CSF IL-10/IL-6 dynamics in DLBCL, SCNSL, and PCNSL, we systematically summarized current literature findings in Table 4 . Distinct from prior PCNSL-focused studies (12–15), our data suggest that serum IL-10/IL-6 ratios can predict CNS relapse risk in asymptomatic DLBCL patients, potentially enabling preemptive therapeutic strategies. Importantly, unlike CSF-based diagnostics, our serum-based approach is minimally invasive and requires no additional sampling. The strong correlation observed between serum and CSF ratios further suggests that systemic immune dysregulation may precede CNS involvement, offering a critical window for early detection.

Our findings demonstrate that an elevated serum IL-10/IL-6 ratio (≥2.30) at initial diagnosis serves as an independent predictor of SCNSL development and disease progression in DLBCL. Incorporating this ratio significantly enhances the prognostic accuracy of both CNS-IPI and NCCN-IPI scoring systems. These results underscore the potential of integrating inflammatory cytokine biomarkers with established clinical indices to refine risk stratification. Nonetheless, clinical translation of these biomarkers will require standardized cutoff definitions, multicenter validation, and longitudinal studies to monitor dynamic changes throughout disease progression and treatment. Further research is essential to address these gaps and realize their full clinical utility.