UC San Diego

Background: Physical activity after a diagnosis of breast cancer is associated multiple positive effects. However, the literature indicates that patients’ activity decreases during chemotherapy. Little is known about when during chemotherapy activity changes, as existing research has relied on self-reported data. Monitoring devices like accelerometers offer more objective measures, nevertheless, these devices may introduce other sources of measurement error.

Methods: Chapters 1 and 2 leverage data from Reach for Health (RFH), a trial that encouraged increased physical activity among breast cancer survivors. At baseline and study completion, participants wore accelerometers and answered the GPAQ. Chapter 1 uses the GPAQ estimates along with accelerometer measures, processed using standard cut-points and a machine learning algorithm, to assess agreement of activity estimates. Comparisons are assessed using mixed effects regression models. Chapter 2 uses a pseudo-simulation to generate missing wear patterns. The simulated and true data is used to compare 6 possible techniques to account for missing accelerometer wear. Chapter 3 analyzed the Activity in Treatment (ACT) study which enrolled 32 women prior to starting chemotherapy for breast cancer, all women were given a Fitbit activity monitor to wear throughout chemotherapy. Restricted cubic splines assessed non-linear patterns of activity.

Results: At baseline, self-report and machine learning provided similar activity estimates; while estimates of activity change were only similar between cut-point and machine learning, the magnitude of agreement with self-report was differential by group. Random slope imputation and an accelerometer specific multiple imputation performed best in correcting for missing wear time. MVPA declined linearly at an average of 1.4 min/day (p=0.002) for every 10% of the duration of chemotherapy that passed, while TPA declined linearly at an average of 13.4 min/day (p=0.0007) for every 10% of chemotherapy that passed. This decline occurred until approximately half way through chemotherapy. Additionally, a HER2+ receptor status was associated with a greater rate of decline in MVPA.

Discussion: Our findings highlight the importance of targeting physical activity interventions during active treatment for breast cancer and increases our ability to standardize research practices regarding the processing and analysis of physical activity data.