Lead Investigator: Sheng Luo, Duke University
Title of Proposal Research: A Novel Machine Learning-Based Prediction Model for Outcomes in Patients with Untreated Diffuse Large B Cell Lymphoma
Vivli Data Request: 8956
Funding Source: None
Potential Conflicts of Interest: None
Summary of the Proposed Research:
Diffuse large B cell lymphoma (DLBCL) is the most common subtype of non-Hodgkin lymphoma (a cancer that starts in the body’s lymphatic system and can spread from there), accounting for approximately 30% of all cases. This aggressive cancer has a varied clinical presentation and response to treatment, with 60-70% of patients still alive after 5 years. Given this variability, accurate prediction of treatment outcomes and patient prognosis is essential for the effective management of DLBCL.
We chose previously untreated participants with DLBCL because they provide a clear and uncontaminated baseline for evaluating the efficacy and safety of new therapies. By studying treatment-naive patients, we can accurately assess how well the treatment works without the confounding effects of prior therapies, which might have altered the disease’s characteristics or induced resistance. Additionally, untreated participants allow us to observe the natural course of the disease and identify early prognostic markers that can guide future therapeutic strategies.
The prognosis of diffuse large B-cell lymphoma (DLBCL) is assessed using the International Prognostic Index (IPI). This index incorporates factors such as age, performance status (PS – a score that estimates the patient’s ability to perform certain activities of daily living), Ann Arbor stage (an indication of how far the lymphoma has spread), serum lactate dehydrogenase (LDH – a chemical found in the blood which can indicate how advanced the lymphoma might be), and extra-nodal involvement (presence of the lymphoma at body sites other than the lymphatic system). Over the years, advances in diagnostic methods and treatment options have led to improved prognoses for DLBCL patients, particularly those in high-risk categories. Consequently, while the IPI maintains its prognostic significance, its capacity to predict treatment failure has diminished. Efforts to enhance the IPI have resulted in modest improvements. A recent study revealed that the National Comprehensive Cancer Network ([NCCN]-IPI) performed best; however, the 5-year overall survival (OS) of the poorest prognostic group was still only 49%.
Recent studies have focused on incorporating medical imaging data (radiomics) derived from positron emission tomography (PET)/computerised tomography (CT) scans into prediction models, some of which have demonstrated significant predictive ability for treatment outcomes. Furthermore, combinations of these radiomics features with clinical variables, such as age, performance status, and Ann Arbor stage, have outperformed the IPI in predicting progression-free survival (PFS) and overall survival (OS).
Meanwhile, gene expression profiling (identifying the genes present in a tumor) has provided significant insight into biological factors underlying divergent clinical outcomes in DLBCL. However, it is still unclear how best to integrate these molecular biomarkers (something found in blood, other body tissues that is a sign of a normal or abnormal process, or of a disease) into the prediction model due to their complexity and the need for specialized tests. More importantly, most recent studies have focused on some specific aspects of DLBCL prognosis, rather than developing comprehensive prediction models that integrate multiple factors using advanced machine-learning techniques.
This project aims to use machine learning techniques to integrate various clinical and biological factors to develop a more accurate and personalized prediction model for patients with previously untreated DLBCL.
Requested Studies:
A Phase III, Multicenter, Open-Label Randomized Trial Comparing the Efficacy of GA101 (RO5072759) in Combination With CHOP (G-CHOP) Versus Rituximab and CHOP (R-CHOP) in Previously Untreated Patients With CD20-Positive Diffuse Large B-Cell Lymphoma (DLBCL)
Data Contributor: Roche
Study ID: NCT01287741
Sponsor ID: BO21005
A Comparative, Randomized, Parallel-group, Multi-center, Phase IIIB Study to Investigate the Efficacy of Subcutaneous (SC) Rituximab Versus Intravenous (IV) Rituximab Both in Combination With CHOP (R-CHOP) in Previously Untreated Patients With CD20-positive Diffuse Large B-cell Lymphoma (DLBCL)
Data Contributor: Roche
Study ID: NCT01649856
Sponsor ID: MO28107
Phase III Randomized Study of R-CHOP V. Dose-Adjusted EPOCH-R With Molecular Profiling in Untreated De Novo Diffuse Large B-Cell Lymphomas
Data Contributor: Project Data Sphere
Study ID: NCT00118209
Sponsor ID: CALGB-50303
A Phase II Study of Epratuzumab, Rituximab (ER)-CHOP for Patients With Previously Untreated Diffuse Large B-Cell Lymphoma
Data Contributor: Project Data Sphere
Study ID: NCT00301821
Sponsor ID: NCCTG-N0489
A Phase III Multicenter, Open-label Study of Rituximab Alternative Dosing Rate in Patients With Previously Untreated Diffuse Large B-cell or Follicular Non-Hodgkin’s Lymphoma
Data Contributor: Roche
Study ID: NCT00719472
Sponsor ID: U4391g
A Randomized, Open-label, Mutli-centre Study to Evaluate Patient Preference With Subcutaneous Administration of Rituximab Versus Intravenous Rituximab in Previously Untreated Patients With CD20+ Diffuse Large B-cell Lymphoma or CD20+ Follicular Non-Hodgkin’s Lymphoma Grades 1, 2, OR 3A
Data Contributor: Roche
Study ID: NCT01724021
Sponsor ID: MO28457