Central Nervous System Relapse in Diffuse Large B Cell Lymphoma—Predictors at Diagnosis

Background Central nervous system (CNS) relapse of diffuse large B cell lymphoma (DLBCL) is an infrequent outcome with very poor prognosis. Initial studies have identified risk factors for CNS relapse and the CNS international prognostic score ((CNS-IPI) (Schmitz et al, JCO 2016) has been validated in various cohorts. Biologic characteristics of the disease are not included in the derivation of this model. We sought to validate this risk factor model in our single center DLBCL population in search for additional disease - specific factors that could further improve prediction of CNS relapse.

Methods The hematologic malignancies database of the University Hospitals Seidman Cancer Center was accessed to identify newly diagnosed DLBCL patients between 2002 and 2014. Patients with extra-nodal CNS involvement at diagnosis were excluded. Through retrospective chart review data, we identified patient and disease characteristics, and outcomes data including response, relapse and death was collected. Patients who underwent therapy and had follow up data were included for final analysis. Univariate Cox proportional hazards analysis was conducted on the risk factors and those reaching statistical significance (p < 0.05) were selected for inclusion in multivariate proportional hazards model to help identify predictors for CNS relapse. CNS IPI score was applied to our population and cumulative incidence (with death as competing risk) was used to estimate the incidence of CNS relapse with the Gray test performed for comparisons between groups.

Results: Five hundred and twenty-four patients were included for final analysis. Median age in our cohort was 64 years, with 312 (59.5 %) patients above the age of 60 years. 322 (63 %) patients had stage (III-IV), and 220 patients (37%) had elevated LDH at diagnosis. Two hundred and eighty five patients had extra-nodal disease involvement at the time of diagnosis with 88 patients involving more than one site. Baseline patient and disease characteristics are listed in Table 1. Four hundred (76%) patients received R CHOP as first line therapy, 395 patients achieved complete remission with first line therapy. At a median follow up of 64 months, 187 patients died and 171 patients had relapsed.

Twenty-one patients developed CNS relapse in the follow up period, at median of 8.2 months after treatment initiation (95% CI 6.2-9.6) months. Applying the CNS-IPI risk scoring among the evaluable patients, 149 patients (36%) were found to be low risk (0-1 factors), 222 patients (53%) were intermediate risk (2-3 factors) and 47 patients (9%) were high risk (4-6 factors) for CNS relapse. Univariate analysis showed advanced stage (stage III-IV), number of extra-nodal sites involved, liver involvement, non- GCB subtype, translocation t (14; 18) and body mass index were significantly associated with CNS relapse (p<0.05). In multivariate analysis, only non- GCB subtype and translocation t (14; 18) (p = 0.023 and 0.028 respectively) were significant. Table 2 summarizes the results of Cox proportional hazards analysis.

We evaluated the cumulative incidence of CNS relapse in the high-risk group, defined as CNS-IPI >1, (n=269, 64%). The 2-year CNS relapse cumulative incidence was at 5.1 % (95% CI: 2.8- 9%) compared to 0% for patients with CNS IPI ≤ 1 (p=0.027) (Figure 1).

Median survival after CNS relapse was 3.6 months (0 - 150.2 months). Overall survival from start of initial therapy was 16 months (95% CI 9.5 -81.9 months) for patients with CNS relapse vs 170 months (95% CI 120.3 months - not reached), p<0.001]. (Figure 2)

Summary

Prognosis of CNS relapsed of DLBCL is very poor. CNS IPI score helps identify a cohort that may benefit from CNS prophylaxis, and may help avoid unnecessary diagnostic and therapeutic intervention in the low risk cohort, i.e., CNS- IPI score <1. Hepatic involvement at diagnosis may be an additional clinical characteristic associated with high risk of CNS relapse. Biological features including non-GCB subtype and translocation t (14; 18) are novel risk factors that should be considered in the development of future risk models.

View largeDownload slide

View largeDownload slide

Close modal