Objective:

To assess the relationship between subcutaneous accelerometer signals and treadmill walking speed and evaluate an insertable cardiac monitor (ICM) algorithm for classifying activity.

Approach:

• Study population: Exercise-substudy participants are adults with New York Heart Association class II or III heart failure enrolled in the multicenter LUX-Dx TRENDS study who are considered suitable for exercise testing.
• Assessments: Participants maintain stationary postures and complete progressively faster treadmill stages. Investigators analyze accelerometer signals and compare algorithm classifications with a prespecified reference standard.
• Analysis: The primary classification analysis defines walking speeds ≥1.2 mph as active and stationary postures as inactive, excluding the 0.7-mph treadmill stage. Agreement is assessed in dataset 2 using Cohen’s kappa.

Key Findings:

• The exercise substudy includes 104 participants at 26 US sites; device data are analyzed for 56 participants in dataset 1 and 45 in dataset 2. Device data are unavailable for 3 participants.
• Activity signals increase by an estimated 33.7 mG per 1-mph increase in walking speed in dataset 1 and 32.3 mG in dataset 2; both associations have P < .0001.
• In dataset 2, agreement between algorithm classifications and the reference standard is κ = 0.971 (95% CI, 0.947–0.985), exceeding the prespecified goal of κ > 0.8.
• Sensitivity is 0.974, specificity is 0.994, positive predictive value is 0.991, and negative predictive value is 0.982.
• The kappa performance goal remains met in sensitivity analyses classifying the 0.7-mph treadmill stage as either active or inactive.

Interpretation:

In this ambulatory heart failure population, subcutaneous accelerometer signals show a strong positive association with treadmill walking speed, and the device algorithm shows high agreement with the prespecified activity reference standard under controlled conditions.

Limitations:

• Testing is limited to controlled treadmill exercise and stationary postures; other activities, including biking and swimming, are not studied.
• Gait differences, treadmill calibration, operator compliance, reference-label inaccuracies, and external vibration may affect results. Several potential influences, including electromagnetic interference, temperature variation, and respiratory rate changes, are not assessed.
• Stationary measurements may not reflect prolonged rest during sleep. Selection may exclude frailer or gait-limited patients, and substantial attrition occurs at higher speeds. Algorithm performance is not evaluated within patient subgroups.

Conclusion:

The findings demonstrate the feasibility of accelerometer-based activity assessment using a miniaturized subcutaneous ICM. Further research should evaluate the algorithm’s real-world utility for detection, risk stratification, and patient management.

Sources:

• Accelerometer-Based Activity Sensing by an Insertable Cardiac Monitor In Ambulatory Patients: A Subanalysis of LUX-Dx TRENDS (LUX-Dx Heart Failure Sensors in an Insertable Cardiac Monitor System Clinical Study)