Digital physiological monitoring for personalized prediction of physical function and fatigue in men with castration sensitive prostate cancer.

124 Background: Androgen deprivation therapy (ADT) associated fatigue and functional impairment presents challenges to patient quality of life (QOL) and treatment adherence. Clinical features and biomarkers predictive of ADT-associated side effects remain elusive. Objective measures of physical function may have the potential to prognosticate side effects for men with prostate cancer, in particular fatigue. Methods: Men with castration sensitive prostate cancer (CSPC), initiating ADT as monotherapy or in combination with an androgen receptor signaling inhibitor (ARSI), enrolled in this prospective observational study. Patients underwent continuous mobility monitoring for 12 months after ADT initiation. Step count, heart rate and sleep were measured via smartwatch. Fatigue severity and QOL were assessed monthly using the FACIT-Fatigue and FACT-P surveys, respectively. Descriptive statistics and linear mixed modeling were used for analysis. Results: 19 CSPC patients were analyzed: mean age 65.7 years (SD 7.7), 78.9% high risk localized/oligometastatic, 94.7% receiving ADT+radiation with or without an ARSI. Total weekly compliance was 91.6% (SD 5.5%). Average daily baseline step-count was 3,570 (SD 1,881) and declined by 23.9% at 12-months. A statistically weak inverse correlation between step count and fatigue severity was observed (R2=0.075). This prompted further exploratory analysis of individual step count as a predictive measure of physical function. To control for high inter-subject variability, a linear mixed model was used. Individual baseline steps were modeled as a random variable with a mean of 2,782 (SD 1,814; CI [1130.15, 4435.07]; p=0.001). Time on ADT, radiation status, HR and sleep duration were included in this model. Mean daily step count was predicted to decrease by 6.22 steps/week (CI [-10.75, -1.68]; p<.001). Weekly radiation was associated with an average decrease of 275.91 steps (CI [-498.18, -53.32]; p=0.015). Daily and nightly HR was associated with a mean linear increase of 110.98 steps/BPM (CI [88.12, 134.08]; p<0.001) and 100.20 steps/BPM (CI [-122.07, -78.47]; p<0.001), respectively. 1 hour of measured sleep was associated with a decrease of 183.69 steps (CI [-261.98, -104.34]; p<0.001). Overall correlation between actual and predicted step count was 0.88 (R2=0.78). Conclusions: Digital activity monitoring is feasible for men with CSPC on ADT. A weak negative correlation was found between FACIT fatigue severity and average daily step count, while our individualized model accurately predicted daily step count at any point during a 1 year ADT period. While further validation with larger sample sizes is needed, personalized activity predictions may have application in the identification of patients at risk for severe functional impairment, demonstrating value in routine clinical care and clinical trials. Clinical trial information: NCT05390827 .