Use of Natural Language Processing to Identify Myeloid/Lymphoid Neoplasm Patients with Fibroblast Growth Factor Receptor 1 Alterations

Background: Myeloid/lymphoid neoplasms (MLN) with fibroblast growth factor receptor 1 (FGFR1) rearrangement are rare hematologic neoplasms with heterogeneous presentation and a poor prognosis. Pemigatinib, a potent, selective inhibitor of FGFR1-3, is the only approved targeted therapy for adults with relapsed/refractory MLN with FGFR1 rearrangement based on results from the phase 2 open-label FIGHT-203 trial (Vannucchi AM, et al. EHA 2024. P1042). Over a 6-year period, only 47 patients with MLN-FGFR1 were enrolled in FIGHT-203, highlighting the rarity of this disease. To identify FGFR1 rearrangements in patients with suspected MLN, initial screening can be done by conventional cytogenetics to detect translocations in chromosome region 8p11, followed by confirmation via break-apart fluorescence in situ hybridization (FISH), next-generation sequencing (NGS), or polymerase chain reaction testing. Currently, limited data exist on the natural history, diagnosis, and testing patterns for FGFR1 rearrangement in the MLN patient population. The objective of this study was to identify probable cases of MLN with any FGFR1 alteration (MLN-FGFR1) by utilizing natural language processing (NLP) to review electronic medical records (EMR) data in the United States.

Methods: This retrospective study included patients tested for FGFR1 alterations during the study period (1/1/2017-6/30/2024). The study population was derived from Norstella real-world data sources, including linked EMR data sourced from over 40 health systems, 450 hospitals, and 3700 ordering facilities; lab data from national and regional reference labs; and open administrative claims data covering 300 million US lives. Patient identification was carried out using NLP of unstructured EMR data to identify those tested for FGFR1 alterations. Key search terms included, but were not limited to: “FGFR1”, “Fibroblast Growth Factor Receptor 1”, and “8p11”. Results were further narrowed to patients with a probable FGFR1 alteration using search terms such as “positive for FGFR1”, “FGFR1 rearrangement”, and “FGFR1 translocation”, and then validated via manual review by clinicians. Patients without evidence of hematologic malignancy at any time during the study period based on ICD-10-CM diagnosis codes were excluded.

Results: The initial NLP-based search identified 17,486 patients who had been tested for the FGFR1 biomarker. Among these patients, 594 were found to have a probable FGFR1 alteration based on search terms indicating positive results. A subsequent manual review by clinicians validated 75 probable cases of MLN with an FGFR1 alteration. Patients had a median (interquartile range) age of 68 (45-76) years at the time of FGFR1 testing, 65% were male, and 52% were Caucasian. The most common pre-test diagnoses included acute myeloblastic leukemia (n=27 [36%]), myelodysplastic/myeloproliferative neoplasms (n=26 [35%]), acute leukemia (n=19 [25%]), and acute lymphoblastic leukemia (n=15 [20%]). Alterations in FGFR1 were primarily detected via NGS (n=41 [55%]) or FISH (n=27 [36%]) testing. Of the 75 probable MLN-FGFR1 cases, 36 (48%) had an FGFR1 amplification, whereas 27 (36%) had an FGFR1 rearrangement. Additional data on patient characteristics and clinical outcomes from this study are forthcoming and will be presented at ASH 2024.

Conclusions: Our study identified a notable number of probable cases of MLN-FGFR1 in the United States, indicating that the prevalence of MLN-FGFR1 may be higher in the real world than historically known. The study findings underscore the importance of appropriate FGFR1 testing for identification of patients with MLN-FGFR1. The study also suggests that leveraging NLP to analyze EMR data can be a viable approach to improve patient identification and outcomes in rare disorders such as MLN-FGFR1.

Verma:Incyte Corporation: Current Employment, Current equity holder in publicly-traded company. Blecker:Incyte Corporation: Current Employment, Current equity holder in publicly-traded company. Edwards:Panalgo LLC, a Norstella Company: Current Employment. Richards:Panalgo LLC, a Norstella Company: Current Employment. Parasuraman:Incyte Corporation: Current Employment, Current equity holder in publicly-traded company. Braunstein:Incyte Corporation: Current Employment, Current equity holder in publicly-traded company. Vachhani:Amgen: Consultancy, Research Funding; MorphoSys: Consultancy; Daiichi Sankyo: Consultancy; Takeda: Research Funding; Abbvie: Consultancy, Research Funding; Astex Pharmaceuticals: Research Funding; Kartos Therapeutics: Research Funding; Gilead/Forty Seven: Research Funding; Constellation Pharmaceuticals: Research Funding; Blueprint Medicines: Consultancy, Research Funding; Novartis: Consultancy; Cogent Biosciences: Consultancy; Incyte: Consultancy, Membership on an entity's Board of Directors or advisory committees, Research Funding, Speakers Bureau; CTI BioPharma Corp (now Sobi): Consultancy, Research Funding; GenenTech: Consultancy; GlaxoSmith Kline: Consultancy; Seattle Genetics: Research Funding; Stemline: Consultancy; Pfizer: Consultancy; Karyopharm: Consultancy.