From Movement to METs: A Validation of ActTrust(R) for Energy Expenditure Estimation and Physical Activity Classification in Young Adults

This study validates the ActTrust(R) accelerometer as a simple, cost-effective tool for estimating energy expenditure and classifying physical activity intensity in young adults by demonstrating its strong correlation with treadmill speed and metabolic equivalents, as well as its high accuracy in predicting activity levels comparable to the widely used ActiGraph(R) GT3X+.

The Gist

Imagine you want to know exactly how much "fuel" your body burns when you walk, run, or sit around. Scientists call this Energy Expenditure. The most accurate way to measure this is like putting a person in a giant, expensive space helmet that analyzes every breath they take (a method called indirect calorimetry). But, obviously, you can't wear a space helmet to the grocery store or while sleeping.

So, researchers use accelerometers—tiny motion sensors found in fitness trackers and smartwatches. These devices count how much you wiggle, shake, and move. The big question is: Does the number of "wiggles" actually tell us how hard we are working?

This paper is about testing a specific, affordable tracker called the ActTrust® to see if it's a reliable "wiggle-counter" for young adults.

Here is the breakdown of their study using some simple analogies:

1. The Setup: The "Treadmill Test Drive"

Think of the researchers as mechanics testing a new car (the ActTrust tracker) against a gold-standard, luxury car (the expensive ActiGraph® tracker and the breath-analysis machine).

• The Drivers: They recruited 56 healthy young adults (34 men, 22 women).
• The Gear: Each person wore two trackers at the same time: one on their hip (like a belt) and one on their wrist (like a watch). They wore both the new ActTrust and the established ActiGraph.
• The Course: They walked and ran on a treadmill at different speeds, starting from a slow stroll, moving to a jog, and ending with a fast run.
• The Goal: To see if the "wiggle counts" from the cheap tracker matched the actual fuel burn measured by the breath machine.

2. The Discovery: "Wiggles" Match the Effort

The results were great news. Just like a car engine revs higher when you press the gas pedal, the trackers counted more "wiggles" as the people ran faster.

• The Correlation: There was a very strong link between how fast they ran and how many wiggles the devices counted.
• The Translation: The researchers built a mathematical "translator" (a linear equation). This translator takes the raw "wiggle count" from the ActTrust and converts it into METs (Metabolic Equivalents).
  • Analogy: Think of METs as "gear levels."
    • Light (Gear 1-3): Sitting or slow walking.
    • Moderate (Gear 4-6): Brisk walking (you can talk, but you're sweating).
    • Vigorous (Gear 7-9): Jogging (talking is hard).
    • Very Vigorous (Gear 10+): Sprinting (you can't talk at all).

3. The "Speed Limits" (Cut-Off Points)

One of the most important things the study did was set speed limits for the trackers. Before this, we didn't know exactly how many "wiggles" per minute meant you were doing "moderate" exercise versus "vigorous" exercise on an ActTrust.

They created a cheat sheet:

• If your ActTrust counts X wiggles per minute, you are in the "Moderate" zone.
• If it counts Y wiggles, you are in the "Vigorous" zone.

They found that the ActTrust works just as well as the expensive ActiGraph, whether it's on your hip or your wrist. It's like finding out that a reliable, budget-friendly smartphone camera takes photos just as good as a professional DSLR for everyday use.

4. Why This Matters: The "24-Hour Movie"

Most fitness trackers are great at counting steps during the day, but they often struggle to tell the difference between "sleeping" and "sitting still."

The ActTrust is special because it was originally designed to track sleep. This study proves it can also track exercise accurately.

• The Analogy: Imagine a movie camera that can film the whole day without stopping.
  • Old way: You had to use Camera A for the day (exercise) and Camera B for the night (sleep), and they didn't always speak the same language.
  • New way: The ActTrust is a single camera that films the whole 24-hour movie. It can tell you when you were sleeping, when you were sitting on the couch, and when you were running, all in one continuous story.

5. The Bottom Line

• It's Affordable: The ActTrust is much cheaper than the big-name trackers, making it possible for scientists and doctors to study large groups of people without breaking the bank.
• It's Accurate: The study showed that the simple math model they built works very well to guess how much energy you burn.
• It's Simple: Unlike complex AI "black boxes" that are hard to understand, their method is a clear, simple formula that anyone can use.

In short: This paper gives us the "user manual" for a cheap, reliable motion tracker. It tells us exactly how to turn raw movement data into a clear picture of how healthy and active a person is, 24 hours a day. This helps doctors and researchers give better advice on how to stay fit and sleep well.

Technical Summary

1. Problem Statement

Accurately estimating physical activity (PA) levels and energy expenditure (EE) is critical for public health monitoring and clinical research. While the "gold standard" for measuring EE is the doubly labelled water (DLW) technique, it is prohibitively expensive and complex for large-scale studies. Indirect calorimetry is the standard for laboratory validation, but accelerometry offers a scalable, cost-effective alternative.

However, significant challenges exist in the field of accelerometry:

• Device Diversity: Different accelerometer models (e.g., ActiGraph® vs. ActTrust®) use different sensors, signal processing algorithms, and sampling rates, leading to non-comparable data.
• Validation Gaps: While the ActiGraph® GT3X+ is widely validated, the ActTrust® (a lower-cost Brazilian device with integrated light and temperature sensors) lacks validation against gold-standard EE measures.
• Lack of Thresholds: There are no established cut-off points to classify PA intensity (light, moderate, vigorous, very vigorous) using ActTrust® data, limiting its utility in research and clinical settings.

2. Methodology

The study employed a controlled laboratory design to validate the ActTrust® against the widely used ActiGraph® GT3X+ and indirect calorimetry.

• Participants: 56 healthy young adults (34 men, 22 women; aged 18–35). Exclusion criteria included musculoskeletal/cardiovascular diseases or medications affecting metabolic rate.
• Devices:
  • ActiGraph® GT3X+: Triaxial accelerometer, 30 Hz sampling, Vector Magnitude (VM) data.
  • ActTrust® (Condor Instruments): Triaxial accelerometer, 25 Hz sampling, Proportional Integration Mode (PIM) data.
  • Placement: Both devices were worn simultaneously on the non-dominant wrist and the hip (dominant side).
• Protocol:
  • Participants underwent a progressive incremental treadmill protocol consisting of 5 conditions (10 minutes each) separated by 5-minute rests:
    1. Resting
    2. Walking at 3 km/h
    3. Walking at 5 km/h
    4. Walking/Running at 7 km/h
    5. Running at 9 km/h
  • Reference Standard: Oxygen uptake ($VO_2$) was measured breath-by-breath using indirect calorimetry (Quark CPET) to calculate Metabolic Equivalents (METs).
• Data Analysis:
  • Preprocessing: Data was aggregated into 1-second epochs. Mean activity counts were calculated from the central 4 minutes of each condition to ensure physiological steady state.
  • Modeling: A General Linear Model (GLM) was developed to estimate METs from movement counts. A square root transformation was applied to both METs and activity counts to address non-linearity.
  • Equation: $MET^{1/2} = \beta_0 + \beta_1 \times Activity^{1/2}$ (adjusted for device and placement).
  • Classification: Cut-off points were derived for four intensity classes based on MET ranges: Light [0–3), Moderate [3–6), Vigorous [6–9), and Very Vigorous [9–∞).
  • Metrics: Performance was evaluated using Sensitivity, Specificity, Balanced Accuracy, and Area Under the Curve (AUC).

3. Key Contributions

• First Validation of ActTrust®: This is the first study to validate the ActTrust® device against indirect calorimetry for energy expenditure estimation and PA intensity classification.
• Derivation of Cut-off Points: The study provides specific movement count thresholds (counts/min) for both hip and wrist placements for both ActTrust® and ActiGraph® GT3X+ devices.
• Linear Modeling Approach: The authors propose a simple, explainable linear regression model (unlike "black-box" neural networks) to estimate METs from raw movement counts, facilitating easier implementation by practitioners.
• 24-Hour Monitoring Potential: By validating a device capable of monitoring sleep and circadian rhythms alongside PA, the study supports the use of ActTrust® for comprehensive 24-hour rest-activity cycle analysis.

4. Key Results

• Correlations: Strong correlations were found between treadmill speed and both METs ($r=0.95$) and movement counts for all device/placement combinations ($r$ ranging from 0.88 to 0.94, all $p < 0.05$).
• Device Differences: ActTrust® devices generated significantly higher activity counts than the GT3X+ at all speeds, confirming the need for device-specific equations.
• Regression Equations: The study established distinct linear equations for each device and placement (e.g., for ActTrust® hip: $MET^{1/2} = 1.107 + 0.0088 \times Activity^{1/2}$).
• Classification Accuracy:
  • Light and Moderate Activity: The model achieved high balanced accuracy (>0.90) and AUC scores (0.98 for Light, 0.63 for Moderate).
  • Vigorous and Very Vigorous: Accuracy was lower but still robust (Balanced Accuracy >0.77; AUC 0.78 and 0.95 respectively).
  • Comparison: The model's performance (AUC 0.98 for light activity) compared favorably to or exceeded previous studies using complex neural network models (e.g., Santos-Lozano et al., which reported AUCs of 0.8 for light activity).
• Cut-off Points: The derived cut-points for the ActTrust® were significantly different from existing GT3X+ reference values, highlighting that generic thresholds cannot be applied across different devices.

5. Significance and Limitations

Significance:

• Cost-Effectiveness: Validates a low-cost alternative to expensive research-grade accelerometers, making large-scale PA studies feasible in resource-limited settings.
• Harmonization: Provides a framework for translating data between different device types and placements, aiding in meta-analyses and longitudinal studies.
• Holistic Health: Enables the simultaneous assessment of sleep, sedentary behavior, and physical activity using a single, validated device, which is crucial for understanding the interplay between these factors in health outcomes.

Limitations:

• Population: The sample consisted only of young, healthy adults; findings may not generalize to children, the elderly, or clinical populations without further validation.
• Activity Types: The protocol was limited to walking and running on a treadmill. It does not account for non-locomotive activities (e.g., cycling, resistance training) or free-living conditions.
• Protocol Design: The fixed progressive order of the treadmill protocol may have introduced order effects (fatigue or warm-up), though this was mitigated by rest periods.

Conclusion:
The study successfully demonstrates that the ActTrust® is a valid, cost-effective tool for estimating energy expenditure and classifying physical activity intensity in young adults. The derived linear models and cut-off points allow researchers to utilize this device for 24-hour monitoring of the rest-activity cycle, bridging the gap between sleep research and physical activity epidemiology.