Evaluation of Technology-Enabled Monitoring of Patient-Reported Outcomes to Detect and Treat Toxic Effects Linked to Immune Checkpoint Inhibitors.

ABSTRACT

Importance Immune checkpoint inhibitors can produce distinct toxic effects that require prompt recognition and timely management. Objective: To develop a technology-enabled, dynamically adaptive protocol that can provide the accurate information needed to inform specific remedies for immune toxic effects in patients treated with immune checkpoint inhibitors. Design, Setting, and Participants: An open-label cohort study was conducted at a single tertiary referral center from September 6, 2019, to September 3, 2020. The median follow-up duration was 63 (interquartile range, 35.5-122) days. Fifty patients with genitourinary cancers treated with immune checkpoint inhibitors were enrolled. Interventions: A fit-for-purpose electronic platform was developed to enable active patient and care team participation. A smartphone application downloaded onto patients' personal mobile devices prompted them to report their symptoms at least 3 times per week. The set of symptoms and associated queries were paired with alert thresholds for symptoms requiring clinical action. Main Outcomes and Measures: The primary end point of this interim analysis was feasibility, as measured by patient and care team adherence, and lack of increase in care team staffing. Operating characteristics were estimated for each symptom alert and used to dynamically adapt the alert thresholds to ensure sensitivity while reducing unnecessary alerts. Results: Of the 50 patients enrolled, 47 had at least 1 follow-up visit and were included in the analysis. Median age was 65 years (range, 37-86), 39 patients (83%) were men, and 39 patients (83%) had metastatic cancer, with the most common being urothelial cell carcinoma and renal cell carcinoma (22 [47%] patients each). After initial onboarding, no further care team training or additional care team staffing was required. Patients had a median study adherence rate of 74% (interquartile range, 60%-86%) and 73% of automated alerts were reviewed within 3 days by the clinic team. Symptoms with the highest positive predictive value for adverse events requiring acute intervention included dizziness (21%), nausea/vomiting (26%), and shortness of breath (14%). The symptoms most likely to result in unnecessary alerts were arthralgia and myalgia, fatigue, and cough. Conclusions and Relevance The findings of this cohort study suggest an acceptable and fiscally sound method can be developed to create a dynamic learning system to detect and manage immune-related toxic effects.