We evaluated disease trajectories of brain tumor patients using passively generalized data streams from smartphone sensors (as in digital phenotyping).

Patients undergoing radiation therapy for primary (glioblastoma, low-grade glioma and meningioma) and metastatic brain tumors, with good performance status and able to use smartphones were invited in the study. Smartphone application was used to continuously collect capture data from smartphone sensors, including accelerometer, and GPS data, that was used to estimate activity index (ratio of physical activity during the day). EORTC quality of life questionnaires were administered every 2 weeks.  Patients were prospectively monitored for up to 3 months or until disease progression (based on clinical and imaging data) or death.  Data was analyzed using polynomal regression.

At total of 93 patients (54 men; median age 58 years; range 25-69 years) were included in the study. Most common diagnoses were glioblastoma, followed by metastatic brain tumors and head-neck tumors. Thirty-three patients experienced disease progression, and 14 patients died. In all patients, activity data using smartphone sensors allowed earlier identification of disease progression when compared to clinical data alone (5 weeks vs 9 weeks). In subgroup analyses, progression was associated with decrease of activity significant in patients with metastatic brain tumors and in patients <50 years of age, patient with poor mental status (ECOG 1-2 vs 0). There was more steep decline in activity at week 8 between patients who died vs those who did not. There was a significant correlation between passively generated data indices and self-reported symptoms of strenuous activities, pain, rest, sleep, weakness, fatigue and overall health (r values >0.5)

Passively generalized data streams from smartphone sensors is promising method for real-time monitoring and earlier identification of progression of primary and metastatic brain tumors.