Voice as a Digital Biomarker of Neurotoxicity in CAR T-Cell Therapy
The purpose of this study is to collect voice recordings and nervous system (neurologic) assessments from people with non-Hodgkin lymphoma (NHL) or multiple myeloma (MM) who are receiving standard treatment with CAR T-cell therapy. Researchers will study whether these voice recordings and assessments are a practical (feasible) way to monitor for immune effector cell-associated neurotoxicity syndrome (ICANS). Feasibility will be measured by tracking how many participants join the study and complete the assessments.
• Documentation of Disease
• o Patients must have pathologically confirmed Diagnosed with non-Hodgkin lymphoma or multiple myeloma.
• Definition of treatment and ability
‣ Scheduled to receive an FDA-approved CAR-T product: Axi-cel, Liso-cel, Tisa-cel, Ide-cel, Cilta-cel, or Brexu-cel.
⁃ Ability to comply with twice daily voice recordings or daily neurologic assessments, as determined by the investigator.
• Age ≥ 18
• ECOG Performance Status of ≤ 2
• Required
‣ Perform twice daily voice recordings using a smartphone.
⁃ Undergo daily neurologic assessments (e.g., ICE score, tremor evaluation).
⁃ Smartphone ownership.
⁃ Sufficient English proficiency to complete structured voice tasks in the study application.
• Comorbid Conditions
‣ No pre-existing neurological conditions that significantly impair speech, including but not limited to severe dysarthria, expressive aphasia, or neurodegenerative disorders.
⁃ No history of significant speech or voice disorders, including laryngeal paralysis, severe dysphonia, or recent vocal cord surgery or radiation to the area.
⁃ No pathology affecting the vocal cords that could interfere with voice analysis, such as vocal cord paralysis, chronic laryngitis, vocal cord nodules, polyps, granulomas, or malignancies.
⁃ No severe hearing impairment that would interfere with voice assessments
• Language o Proficiency in spoken English is required, without the need for native-level fluency. This ensures participants can accurately perform structured voice tasks, as the application and underlying acoustic models are currently validated only in English, despite the limitation in generalizability. Participants with language barriers that prevent reliabletask completion or data interpretation will be excluded.