The Use of Biomarkers to Predict CNS Involvement in Diffuse Large B-Cell Lymphoma: a Danish Nationwide Registry Study

Diffuse Large B-cell Lymphoma (DLBCL) is a malignant, aggressive lymphoid cancer. The incidence in Denmark is approximately 450 cases per year. In 2/3 of the cases, complete remission is achieved with immunochemotherapy. The remaining 30% will experience relapse and in 5 % of the patients, this will occur in the central nervous system (CNS). CNS relapse has a very poor prognosis with an overall survival of 3-6 months. In order to identify patients at risk of CNS relapse, the CNS-IPI score is used to stratify the patients into three risk groups according to number of risk factors (low 0-1, middle 2-3 and high risk 4-6 which corresponds to 2-year CNS relapse rates of 0,6%, 3,4% and 10,2% respectively). DLBCL can be subdivided by gene expression analysis into three different types based on the cell of origin (ie the stage of the equivalent normal cell development from which the disease arises): the germinal center B-cell (GCB)-like subtype, the activated B-cell (ABC)-like subtype and unclassifiable. The subdivision is of prognostic importance as patients with GCB-like subtype have a 5-year OS of 76% vs 34% in the non-GCB group. Furthermore, studies have found a higher risk of CNS relapse in the ABC-like subtype compared to the GCB subtype0. Other gene rearrangements of potential importance to the risk of CNS relapse is "double hit" (DHL) (5-10% of newly diagnosed DLBCL patients) and MYC/BCL2 co-expressors (double expressors, DEL). Chemotherapeutic CNS prophylaxis is recommended based on the CNS-IPI stratification for the high-risk group (CNS-IPI 4-5) due to an estimated risk of CNS relapse of 10,2%. However, a relapse risk with a specificity of 10,2% results in almost 90% of the patients potentially receiving 'unnecessary' prophylactic chemotherapy with toxic side effects. One study published on data from the GOYA-trial have integrated COO into the CNS-IPI and found an increased sensitivity with a two year relapse risk of 15,2% in the high risk group. In this study we aim to validate the CNS-IPI and evaluate whether the addition of biomarkers for cell of origin (COO) and double hit (DH) DLBCL improves the prediction of later CNS relapse. This will be done through analysis of patientdata from the Danish nationwide lymphoma database, LYFO, on all patients with DLBCL diagnosed from 1.1.2014 to 1.1.2021 combined with pathology reports from the Danish Pathology registry.

Background Diffuse Large B-cell Lymphoma (DLBCL) is a malignant, aggressive lymphoid cancer. The incidence in Denmark is approximately 450 cases per year. In 2/3 of the cases, complete remission is achieved with immunochemotherapy. The remaining 30% will experience relapse and in 5 % of the patients, this will occur in the central nervous system (CNS). CNS relapse has a very poor prognosis with an overall survival of 3-6 months1. In order to identify patients at risk of CNS relapse, the IPI score has been further elaborated and validated using real life data2 to the CNS-IPI score including the five IPI criteria (age greater than 60 years, stage III or IV disease, elevated serum LDH, ECOG performance status \>1, more than 1 extranodal site) with the addition of disease localized to the adrenal glands or the kidneys. The CNS-IPI stratifies the patients into three risk groups according to number of risk factors (low 0-1, middle 2-3 and high risk 4-6 which corresponds to 2-year CNS relapse rates of 0,6%, 3,4% and 10,2% respectively). CNS prophylaxis can be administered by two different approaches; either injected intrathecally (IT) or intravenously (termed high dose, (HD)) with drugs with a well established ability to penetrate the blood-brain-barrier. Recent studies have raised concerns regarding the efficacy of CNS prophylaxis, and current evidence is problematic in relation to: In general the studies are small and limited by the low event rate of CNS relapses3 4 * Patient selection (ie. patients, who receive prophylaxis per se have a higher risk of CNS relapse than the background population) which result in a higher rate of CNS relapse in the patients receiving prophylaxis * Prophylaxis given by clinician's choice5 * Some register based studies cannot distinguish between the timing or type of prophylaxis administered1 However recent prospective studies with focus on HD prophylaxis have shown a lower than expected number of CNS events67. Furthermore the question has been raised as to whether the timing of the administration matters based on the fact that most relapses occur within the first 6 months after treatment and thus may imply occult CNS-involvement at the time of diagnosis6. In the recent Danish guidelines, administration of prophylactic IV HD MTX is recommended early in the systemic treatment. However due to toxicity, older patients or patients with renal failure must receive either dosage reduced HD MTX or IT MTX. Cell of origin DLBCL can be subdivided by gene expression analysis into three different types based on the cell of origin (ie the stage of the equivalent normal cell development from which the disease arises): the germinal center B-cell (GCB)-like subtype, the activated B-cell (ABC)-like subtype and unclassifiable8. In the clinical setting, gene expression profiling is not commonly used, but the subdivision can be approximated by immunohistochemistry where two subgroups can be identified as GCB and non-GCB by the markers CD10, BCL6 and MUM1. The subdivision is of prognostic importance as patients with GCB-like subtype have a 5-year OS of 76% vs 34% in the non-GCB group9. Furthermore, studies have found a higher risk of CNS relapse in the ABC-like subtype compared to the GCB subtype10. As previously described, CNS-IPI stratifies patients into a high-risk group (CNS-IPI 4-5) with a risk of CNS relapse of 10,2% and administration of CNS prophylaxis is recommended based on this stratification. However, a relapse risk with a specificity of 10,2% results in almost 90% of the patients potentially receiving 'unnecessary' prophylactic chemotherapy with toxic side effects. The need for better stratification tools is obvious. One study published on data from the GOYA-trial have integrated COO into the CNS-IPI and found an increased sensitivity with a two year relapse risk of 15,2% in the high risk group11. Double hit/triple hit lymphomas A double hit (DHL) is present in approximately 5-10% of newly diagnosed DLBCL patients with rearrangements of the proto-oncogenes MYC and BCL2 and/or BCL6. A larger group of newly diagnosed DLBCL patients (30-40 %)are MYC/BCL2 co-expressors (double expressors, DEL)12. Patients with MYC and BCL2 and/or BCL 6 transformation are known to have a worse OS with a poorer response on standard treatment with R-CHOP13. Regarding the risk of CNS relapse, one study on DHL showed an increased risk of CNS relapse with a HR 2,14 over a three year follow-up period14 These findings are in accordance with previous, but smaller, studies15 16. Another study found that DEL alone was an independent risk of CNS relapse. In Cox regression multivariate analysis including the COO (IHC), IPI group and MYC/BCL2 (IHC), only the IPI (HR 2.18, P=0.02) and MYC+BCL2+ IHC (HR=3.76, P=0.007) were associated with an increased risk of CNS relapse.17 However others have not been able to confirm this association.11 Aims The primary aim of this retrospective, register-based cohort study is to validate the CNS-IPI and evaluate whether the addition of biomarkers for cell of origin (COO) and double hit (DH) DLBCL improves the prediction of later CNS relapse. Furthermore, we will explore the predictive value of the CNS-IPI in regard to extent and localization of CNS relapse and investigate the correlation between CNS relapse and primary extranodal manifestations. Finally, we aim to evaluate the effect of the type and timing of CNS prophylaxis. Perspectives The dismal prognosis of CNS relapse is an unsolved issue for patients and clinicians worldwide. The toxicity of the CNS-prophylaxis, the low sensitivity and specificity of the CNS-IPI and the conflicting results from clinical studies are contributing factors to the lack of consensus regarding administration of as well as type of prophylaxis. The need for better evidence to guide the clinical practice is obvious. Despite the inherent limitations of a retrospective study, real-world data from a large population based cohort using the LYFO and PatoBank databases is expected to further clarify the prognostic effect of biomarkers as well as the best possible clinical practice. Methods Study design A retrospective multi center cohort study with data collection from all five Danish regions (Capital City Region, Region Zealand, Southern Denmark, Mid Jutland and North Jutland) by local investigators. From the Danish nationwide lymphoma database, LYFO, data will be collected on all patients with DLBCL diagnosed from 1.1.2014 to 1.1.2021. This will result in an estimated total of 2969 patients equivalent to the incidence of 424 cases per year. On the patients with CNS relapse a medical record review will be performed with focus on site of relapse and type of prophylaxis (if administered). Furthermore a fusion of data from the Danish Pathology registry will provide information on performed ICH and FISH analyses. Data collection Data will be registered in RedCap by the data collectors. RedCap is an electronic, secure data collection platform approved by the Danish Data Protection Agency to store sensitive patient information. The local/regional data collectors will access a folder in RedCap with CPR-numbers from patients in their own department or region alone and the CRFs will be completed directly in RedCap. Access to the medical records is granted by the head of department at the different sites of data collection. Statistics The primary outcome will be analyzed using survival statistics. Kaplan-Meier curves will be used for describing time to CNS relapse and risk of CNS relapse. Cox proportional hazard models will be used to evaluate the prognostic value of CNS-IPI, COO and DH, adjusted for other measures. Secondary outcomes mainly consist of subgroup analyses and will be analyzed using similar methods as the primary outcome. Overall survival will be analyzed using Kaplan-Meier statistics and Cox-proportional hazard models. Furthermore, descriptive statistics will be performed to describe the study population. Statistic calculations will be performed using SAS, R or similar. Results for primary and secondary outcomes will be presented with 95% confidence intervals. All P-values will be two-sided and p-values below 0.05 will be considered statistical significant. Ethics The study has achieved approval from the regional Data Protection Agency and The Danish Patient Safety Authorities. Access to the data in PatoBank has in addition achieved approval from the national Data Protection Agency. The study does not require direct patient contact so approval from the regional ethics committee is not necessary. This study is retrospective and therefore cannot influence any course of treatment or outcome. Reference list 1. El-Galaly, T. C. et al. The number of extranodal sites assessed by PET/CT scan is a powerful predictor of CNS relapse for patients with diffuse large B-cell lymphoma: An international multicenter study of 1532 patients treated with chemoimmunotherapy. Eur. J. Cancer 75, 195-203 (2017). 2. Schmitz, N. et al. CNS International Prognostic Index: A risk model for CNS relapse in patients with diffuse large B-Cell lymphoma treated with R-CHOP. J. Clin. Oncol. 34, 3150-3156 (2016). 3. Tai, W. M. et al. Central nervous system (CNS) relapse in diffuse large B cell lymphoma (DLBCL): Pre- and post-rituximab. Ann. Hematol. 90, 809-818 (2011). 4. Qualls, D. \& Abramson, J. S. Advances in risk assessment and prophylaxis for central nervous system relapse in diffuse large B-cell lymphoma. Haematologica 104, 25-34 (2019). 5. Eyre, T. A., Djebbari, F., Kirkwood, A. A. \& Collins, G. P. A systematic review of the efficacy of CNS prophylaxis with stand-alone intrathecal chemotherapy in diffuse large B cell lymphoma patients treated with anthracycline-based chemotherapy in the rituximab era. Haematologica haematol.2019.229948 (2019) doi:10.3324/haematol.2019.229948. 6. Holte, H. et al. Dose-densified chemoimmunotherapy followed by systemic central nervous system prophylaxis for younger high-risk diffuse large B-cell/follicular grade 3 lymphoma patients: Results of a phase II Nordic lymphoma group study. Ann. Oncol. 24, 1385-1392 (2013). 7. Ferreri, A. J. M. et al. Risk-tailored CNS prophylaxis in a mono-institutional series of 200 patients with diffuse large B-cell lymphoma treated in the rituximab era. Br. J. Haematol. 168, 654-662 (2015). 8. Alizadeh, A. A. et al. Distinct types of diffuse large B-cell lymphoma identified by gene expression profiling. Nature 403, 503-511 (2000). 9. Hans, C. P. et al. Confirmation of the molecular classification of diffuse large B-cell lymphoma by immunohistochemistry using a tissue microarray. Blood 103, 275-282 (2004). 10. Savage, K. J. et al. Impact of dual expression of MYC and BCL2 by immunohistochemistry on the risk of CNS relapse in DLBCL. Blood 127, 2182-2188 (2016). 11. Klanova, M. et al. Integration of cell of origin into the clinical CNS International Prognostic Index improves CNS relapse prediction in DLBCL. Blood 133, 919-926 (2019). 12. Cheah, C. Y., Oki, Y., Westin, J. R. \& Turturro, F. A clinician's guide to double hit lymphomas. Br. J. Haematol. 168, 784-795 (2015). 13. Landsburg, D. J. et al. Outcomes of patients with double-hit lymphoma who achieve first complete remission. J. Clin. Oncol. 35, 2260-2267 (2017). 14. Oki, Y. et al. Double hit lymphoma: The MD Anderson Cancer Center clinical experience. Br. J. Haematol. 166, 891-901 (2014). 15. Snuderl, M. et al. Rearrangements Are Aggressive Neoplasms with Clinical and. Am. J. Surg. Pathol. 34, 327-340 (2010). 16. Le Gouill, S. et al. The clinical presentation and prognosis of diffuse large B-cell lymphoma with t(14;18) and 8q24/c-MYC rearrangement. Haematologica 92, 1335-1342 (2007). 17. Savage. No Title. https://ashpublications.org/blood/article/124/21/4 (2014).