2240 - Smart Surveillance: Digital Mobility Biomarkers from Wearables and Smartphones for Early Detection of Brain Metastases in High-Risk Cancer Patients

Purpose/Objective(s): Brain metastases (BM) frequently present with advanced neurologic symptoms requiring urgent radiation therapy. We hypothesized that continuously captured digital mobility metrics demonstrate measurable change proximal to MRI-confirmed BM diagnosis.

Materials/Methods:

We retrospectively analyzed 157 patients with high-risk solid tumors enrolled in continuous passive digital monitoring at one institution. Daily step count and stride metrics were temporally aligned to MRI-confirmed BM diagnosis. Data were cleaned for temporal continuity and device adherence. For cohort-level analysis, we required =7 days of wear in both pre- and post-diagnostic windows within ±90 days of BM diagnosis. Baseline mobility was defined as median daily step count in the pre-diagnostic window, and compared with post-diagnostic values using paired testing. In a representative high-adherence vignette, segmented interrupted time series (ITS) regression modeled step count as a function of time, diagnostic event, and post-event slope, with heteroscedasticity- and autocorrelation-robust standard errors.

Results:

Step count data were available for 157 patients and stride length for 149. Thirty-three patients had radiology-defined BM, including 26 MRI-confirmed intracranial events; lung cancer predominated (92.3%).

Among patients meeting wear criteria (n=19), daily step count declined significantly following BM diagnosis relative to baseline (paired analysis, p<0.05).

In the representative patient, weekly paired medians demonstrated significant mobility decline immediately following diagnosis (Wilcoxon p=0.039). ITS modeling revealed a significant pre-diagnostic negative slope (-87 steps/day, p<0.001), followed by an immediate level decrease of -4,535 steps at diagnosis (p<0.001), representing approximately 30% reduction from baseline activity. A significant post-diagnostic slope change (+255 steps/day, p<0.001) suggested partial functional recovery after intervention.

Across evaluable patients, directional mobility decline proximal to diagnosis was observed, though magnitude varied by individual trajectory.

Conclusion:

In this pilot cohort, wearable-derived mobility data demonstrated statistically significant inflection at MRI-confirmed BM diagnosis. Absolute step count reductions exceeded 4,000 steps in representative cases, indicating clinically meaningful functional decline. These findings support the feasibility of passive digital mobility monitoring as an objective adjunct to neurologic surveillance in high-risk oncology populations. Prospective validation is ongoing through BEACON (Biometrics-Enabled Analytics for Cancer Observation and Notification), evaluating continuous digital performance status and adaptive clinical intervention.