DEVELOPMENT OF A NOVEL ALGORITHM TO IDENTIFY CAPILLARY LEAK SYNDROME (CLS) IN US CLAIMS DATA
Author(s)
Shai Shimony, MD, MPH1, Naveen Pemmaraju, MD2, Gabriel N. Mannis, MD3, Eunice S. Wang, MD4, Thomas W. LeBlanc, MD5, Brittany Umer, PhD6, Adrian Mui, MS6, Mansoure Jahanian, MS6, Alessandra Tosolini, BS7, John Katsetos, PA7, Anthony S. Stein, MD8, Andrew A. Lane, MD, PhD1;
1Dana-Farber Cancer Institute, Boston, MA, USA, 2The University of Texas MD Anderson Cancer Center, Houston, TX, USA, 3Stanford University, Stanford, CA, USA, 4Roswell Park Comprehensive Cancer Center, Buffalo, NY, USA, 5Duke University School of Medicine, Durham, NC, USA, 6Klick Health, Toronto, ON, Canada, 7Menarini Group, New York, NY, USA, 8City of Hope, Duarte, CA, USA
1Dana-Farber Cancer Institute, Boston, MA, USA, 2The University of Texas MD Anderson Cancer Center, Houston, TX, USA, 3Stanford University, Stanford, CA, USA, 4Roswell Park Comprehensive Cancer Center, Buffalo, NY, USA, 5Duke University School of Medicine, Durham, NC, USA, 6Klick Health, Toronto, ON, Canada, 7Menarini Group, New York, NY, USA, 8City of Hope, Duarte, CA, USA
OBJECTIVES: CLS is a systemic inflammatory adverse event (AE) associated with several conditions and medications, including anticancer therapies. Identifying CLS in claims data is challenging due to the lack of a specific CLS ICD-10 code and overlap of CLS with other clinical diagnoses (eg, sepsis, hypersensitivity reactions). We developed an algorithm for identifying patients who experienced CLS in a US claims database.
METHODS: The algorithm identified potential candidate CLS-causing drugs based upon product label, reports in scientific data, or documentation in the FDA AE Reporting System database. Patients from the PurpleLab claims database were included if, within 2 weeks of initiating candidate drug treatment, they had a diagnosis of diseases of the capillaries (ICD-10: I78.8/9); or had abnormal albumin levels or received intravenous albumin; or had ≥3 indicators of fluid resuscitation/management, edema, hypotension, infusion-related reactions, cardiac arrest, or cardiopulmonary failure. Key exclusion criteria included diagnosis of sepsis, severe hypersensitivity, cytokine release syndrome, or prior history of idiopathic CLS.
RESULTS: Eighty-one CLS-causing drugs were identified in the evaluation of ~7.5 million patients. After incorporating inclusion and exclusion criteria, CLS was not observed in 4,560,767 patients, and CLS was observed in 14,091 patients. The most common inclusion criteria combination was edema plus hypotension plus other indicators; sepsis was the most common exclusion criterion. The algorithm identified CLS incidence rates of 21% in tagraxofusp-, 0.46% in filgrastim-, and 0.70% in gemcitabine-treated patients. For comparison, the investigator-assessed CLS rate in the pivotal tagraxofusp study was 21% (Pemmaraju JCO 2022); filgrastim and gemcitabine have reported CLS rates of <1% (Izzedine Kidney Int Rep 2022).
CONCLUSIONS: We generated a novel claims-based algorithm for CLS incidence with rates similar to known CLS-causing drugs including tagraxofusp, filgrastim, and gemcitabine. This claims-based algorithm will be validated in future evaluations to identify predictors of CLS in tagraxofusp-treated patients.
METHODS: The algorithm identified potential candidate CLS-causing drugs based upon product label, reports in scientific data, or documentation in the FDA AE Reporting System database. Patients from the PurpleLab claims database were included if, within 2 weeks of initiating candidate drug treatment, they had a diagnosis of diseases of the capillaries (ICD-10: I78.8/9); or had abnormal albumin levels or received intravenous albumin; or had ≥3 indicators of fluid resuscitation/management, edema, hypotension, infusion-related reactions, cardiac arrest, or cardiopulmonary failure. Key exclusion criteria included diagnosis of sepsis, severe hypersensitivity, cytokine release syndrome, or prior history of idiopathic CLS.
RESULTS: Eighty-one CLS-causing drugs were identified in the evaluation of ~7.5 million patients. After incorporating inclusion and exclusion criteria, CLS was not observed in 4,560,767 patients, and CLS was observed in 14,091 patients. The most common inclusion criteria combination was edema plus hypotension plus other indicators; sepsis was the most common exclusion criterion. The algorithm identified CLS incidence rates of 21% in tagraxofusp-, 0.46% in filgrastim-, and 0.70% in gemcitabine-treated patients. For comparison, the investigator-assessed CLS rate in the pivotal tagraxofusp study was 21% (Pemmaraju JCO 2022); filgrastim and gemcitabine have reported CLS rates of <1% (Izzedine Kidney Int Rep 2022).
CONCLUSIONS: We generated a novel claims-based algorithm for CLS incidence with rates similar to known CLS-causing drugs including tagraxofusp, filgrastim, and gemcitabine. This claims-based algorithm will be validated in future evaluations to identify predictors of CLS in tagraxofusp-treated patients.
Conference/Value in Health Info
2026-05, ISPOR 2026, Philadelphia, PA, USA
Value in Health, Volume 29, Issue S6
Code
MSR120
Topic
Methodological & Statistical Research
Topic Subcategory
Artificial Intelligence, Machine Learning, Predictive Analytics, Missing Data
Disease
SDC: Oncology