Modeling Multi-Source Data in Hodgkin Lymphoma: Risks and Outcomes of Relapse/Refractory Disease

Hodgkin lymphoma (HL) is a type of blood cancer that affects the lymphatic system, which is part of the immune system. The lymphatic system is a network of vessels and glands spread throughout your body. In Hodgkin lymphoma, B-lymphocytes (a particular type of white blood cell) start to multiply in an abnormal way and begin to collect in certain parts of the lymphatic system. The affected lymphocytes lose their infection-fighting properties, making you more vulnerable to infection. HL is most often diagnosed in early adulthood or late adulthood, but overall is a rare form of cancer. HL is a highly curable cancer with overall survival of 75-80%. However, survivors are susceptible to severe therapy-induced late effects (LE) that can emerge soon after treatment or decades later, and escalate exponentially over time. LE include secondary cancers, cardiovascular dysfunction, endocrine issues like thyroid dysfunction, infertility, fatigue and lung and nerve damage, and psychological impacts such as anxiety, depression, memory problems, and sleep disturbances. There is no current way to estimate the level of LEs among treatment options or the trade-off between initial disease control vs later toxicity.

In 2018, we formed an international consortium, HoLISTIC (Hodgkin Lymphoma International STudy for Individual Care), now with 80+ members. Applying established data science methods, we created a common data model with a data dictionary (i.e., a comprehensive document that includes the name of each data variable, format of the variables and notes) across all sources, resulting in the creation of an annotated (labelled) database (which enables researchers to aggregate (combine) data in a standardized way) of more than 30,000 adult HL patients from clinical trials and prominent registries. Importantly, this robust database was designed to incorporate new study results and new agents, as they become available. This is highly relevant, given the promising results of novel agents—first used in the relapsed/refractory (R/R) disease and now in the frontline setting for initial treatment (1L).

Our parent grant awarded in March 2022, proposed the synthesis of rich, multi-source data to accomplish some specific aims: development of pre-treatment, clinical prediction models, using baseline factors; multi-state models (which are statistical frameworks used to study how individuals move between different clinical states or junctures over time, such as from treatment start to interim disease response via Positron Emission Tomography (PET) scan, or from interim PET scan to the end of planned treatment) to estimate treatment outcomes, based on treatment exposures and treatment response; and simulation models to estimate the probability of late effects, based on cumulative treatment exposure.

While the initial focus of our work was on 1L therapy, we now have the unparalleled opportunity to broaden our consortium to include several new lymphoma experts who are dedicated to understanding risks and outcomes following R/R disease. Through this collaboration, we will expand our database to include nearly 2,400 individual patients who have been treated for R/R disease, prior to and following the introduction of novel agents.

The proposed research will be impactful, as it will more completely characterize treatment exposures and outcomes for patients with HL, by incorporating detailed information about treatment failure in the contemporary era. It will also serve as a powerful proof of principle by highlighting how this multi-source database can be dynamically updated to include new information as it becomes available.