Lead Investigator: Jie Zha, Xiamen University
Title of Proposal Research: Retrospective cohort study with internal and external validation of a predictive tool for progression of disease within 2 years (POD24) risk assessment in follicular lymphoma patients
Vivli Data Request: 7858
Funding Source: National Natural Science Foundation of China
Potential Conflicts of Interest: None
Summary of the Proposed Research:
Follicular lymphoma (FL) is a slow-growing cancer of the immune system that affects white blood cells called lymphocytes.. Based on histology FL is graded as 1 – 3 (low grade to high grade). Treatment patterns for patients with follicular lymphoma grade 3a (FL3a) remain controversial now. Currently, National Comprehensive Cancer Network (NCCN) guidelines recommended some patients are treated as FL and others are treated as Diffuse large B cell lymphoma (DLBCL), which is an aggressive, fast-growing lymphoma and is the most common type of blood cancer. However, which patients should be treated as FL or DLBCL remains poorly defined. Therefore, it is important to perform risk stratification for those patients to guide therapy decisions for FL3a. Using a novel machine learning algorithm, we constructed a risk stratification tool mainly based on the clinical features in the derivation cohort of 543 FL3a patients and compared with conventional risk models. This machine learning model was developed to discriminate FL3a patients into low- and high-risk groups, which will help clinicians select individual treatment strategies.
Statistical Analysis Plan:
We randomly divide a large cohort of Chinese FL3a patients into training (75%) and internal validation (25%) cohorts. In the training cohort, we performe traditional univariable and multivariable analyses for PFS and OS using stepwise selection. The patients with missing values in the training dataset will be excluded in the univariable analyses. If a variable met the predetermined significance threshold (P<0.1) in univariable analysis, it would enter further multi-variable Cox regression analysis.Variables with the missing data will be imputed using multivariate imputation by chained equations algorithm before multiple modeling.The selected clinical features entered into multi-tree XGBoost and they will be ranked by this model based on the values for predicting POD24.Then a scoring model based on weighting of the significant variates is constructed. The performance of the models will be evaluated in the internal and external cohorts. For external validation, we expect to use the data set from the clinical trails: SABRINA: NCT01200758 and GALLIUM: NCT01332968.
Requested Studies:
A Multicenter, Phase III, Open-Label, Randomized Study in Previously Untreated Patients With Advanced Indolent Non-Hodgkin’s Lymphoma Evaluating the Benefit of GA101 (RO5072759) Plus Chemotherapy Compared With Rituximab Plus Chemotherapy Followed by GA101 or Rituximab Maintenance Therapy in Responders
Data Contributor: Roche
Study ID: NCT01332968
Sponsor ID: BO21223
A Two-Stage Phase III, International, Multi-Center, Randomized, Controlled, Open-Label Study to Investigate the Pharmacokinetics, Efficacy and Safety of Rituximab SC in Combination With CHOP or CVP Versus Rituximab IV in Combination With CHOP or CVP in Patients With Previously Untreated Follicular Lymphoma Followed by Maintenance Treatment With Either Rituximab SC or Rituximab IV
Data Contributor: Roche
Study ID: NCT01200758
Sponsor ID: BO22334
Public disclosures:
Zha, J., Chen, Q., Zhang, W., Jing, H., Ye, J., Yu, H., Yi, S., Li, C., Zheng, Z., Xu, W. and Li, Z., 2023. Development and Validation of a Machine-Learning Model to Predict POD24 Risk of Follicular Lymphoma. Blood, 142(Supplement 1), pp.3048-3048. Doi : 10.1182/blood-2023-182812