Diagnosis-to-Treatment Interval Is an Important Prognostic Factor with a Time-Dependent Effect Predicting Event-Free Survival after 12 Months from First-Line Treatment in Newly Diagnosed Diffuse Large B-Cell Primary CNS Lymphoma

Background: Diagnostic-to-treatment interval (DTI) was recently described as a strong prognostic factor for newly diagnosed systemic diffuse large B-cell lymphoma patients (DLBCL) with an improvement of event-free survival (EFS) for patients with a longer DTI. These results have some implications for patient selection and result interpretations in clinical trials. This association has not been previously evaluated in DLBCL Primary CNS Lymphoma (PCNSL) patients who also present a clinically aggressive disease.

Patients and Methods: The cohort constited of all consecutive DLBCL PCNSL patients treated in two Hematology Departments of the University of Lyon between 1984 and 2018 (N=244). All patients had DLBCL histology at diagnosis, obtained by brain biopsy (N=235, 96%), vitrectomy (N=4, 2%) or CSF evaluation (N=5, 2%). As first line treatments, all patients but 5 (2%) received high-dose (HD) methotrexate-based chemotherapy, associated with intra-venous rituximab for 154 patients (63%) and HD cytarabine for 182 patients (75%). Consolidation treatment by whole-brain radiotherapy was performed in seventy-six patients (31%). DTI was defined as the number of days between the date of diagnosis (i.e. biopsy) and the date of treatment initiation. Association between DTI and patient characteristics was assessed by chi-square tests or Student t-tests. EFS was defined from the start of therapy to progression, relapse, or death from any cause. As we previously described, prognostic factors such as age and performance status (PS) demonstrate a time-dependent effect on overall survival (OS) in PCNSL limiting the validity of traditional Cox proportional hazard models. We thus used a piecewise Cox model to allow assessment of prognostic effect over different time periods. All survival analyses were done in univariate and multivariate settings and stratified on rituximab use during first-line therapy.

Results: With a median follow-up of 73.5 months, the 5-year EFS and OS rates were 31.6% and 48.4% for whole cohort, respectively. Median DTI was 16 days (range, 1 to 67 days). Short DTI (≤16 days) was associated with a poor PS (ECOG PS 2-4, 53.3% versus 38.5%), altered Karnofsky score (<70%, 49.2% versus 32.0%) but not median of age (63 versus 64 years), CSF level, deep brain involvement, type of symptoms at diagnosis, diagnosis modality, treatment period, inclusion in clinical trial and rituximab used. However, patients with elevated LDH level presented more frequently a longer DTI (41.8% versus 32.0%). DTI was associated with MSKCC risk score but not IELSG score. For prognostic analyses, 205 patients were considered (39 patients with missing data [LDH or PS]). The 5-year EFS rates were 24.6% versus 39.4% for patients with DTI ≤ and > 16 days, respectively ( Figure 1). Using a standard cox model, in univariate analyses, PS (2-4 vs. 0-1) (HR: 1.40, 95%CI, 1.00 – 1.94, P=0.045) and age (per 10-year increase) (HR: 1.19, 95%CI, 1.05-1.36, P=0.006) were associated with EFS but not DTI (≤ or > 16 days) (HR: 0.80, P=0.19), deep involvement (HR: 0.97, P=0.85) and LDH level (HR=1.02, P=0.91). In multivariate analysis, only age was associated with EFS (HR: 1.27, 95%CI, 1.11-1.46, P=0.001). Using a piecewise Cox model over two periods of time (before and after 12 months), we confirmed in multivariate analyses, the time varying effect of PS and age on EFS with a high-risk period before 12 months and no prognostic effect after 12 months (Table 1). We also observed a time-dependent effect for DTI as shown by a significant interaction with time (P=0.02) (Table 1). Indeed, longer DTI was not associated with EFS before 12 months (HR: 1.14, 95%CI, 0.74-1.76, P=0.56) however, it had a strong protective effect after 12 months (HR: 0.44, 95%CI, 0.24-0.86, P=0.02).

Conclusions: In this large cohort of DLBCL PCNSL, a short DTI was mainly associated with poor PS at diagnosis. We confirmed that prognostic factors for PCNSL outcome such as age and PS had time-varying effects with a good predictability of EFS only before 12 months. However, DTI allows prediction of long-term EFS (>12 months) after first line treatment. These results could be related to different biological patterns of tumor aggressiveness. If confirmed in independent PCNSL cohorts, DTI should also be taken in consideration for patient selection and the interpretation of clinical trial results especially for long-term outcome.