• accuracy
• COMSUBIN
• metabolic gases
• monitoring physical activity
• validation
• wearable devices

Purpose - In recent years, interest in monitoring physiological performance using wearable devices has grown exponentially. In particular, for special forces operators, monitoring these parameters is of fundamental importance both during training and in the field, as it enables them to better control and direct their preparation, serving to support the significant efforts required during operational missions. Smartwatches now offer real-time monitoring of key physiological parameters such as heart rate, stress levels, caloric expenditure, and maximal oxygen uptake (VO₂ max). The ability to estimate VO₂ max non-invasively and in everyday environments presents a major opportunity for preventive medicine, athletic performance assessment, and general health monitoring. This study is aimed at comparing VO₂ max values obtained using two commercially available smartwatches (Garmin Descent Mk2 and Garmin Fenix 5) with those derived from the COSMED K5, a portable metabolic analyser considered the reference standard. The objective was to determine whether smartwatches represent a valid alternative to the reference-standard method for estimating maximum oxygen uptake.
Methods - The study involved twenty operators of the Italian Navy Special Forces and Divers Command (COMSUBIN) and was structured in two phases, laboratory- and field-based phase. In the laboratory phase, VO₂ max was measured using the COSMED K5 portable metabolic system during a graded treadmill test (starting at 8 km/h, +1 km/h per minute, 1% incline, to exhaustion), with participants wearing the K5 in backpack configuration and a sealed face mask. The field phase collected smartwatch VO₂ max estimates under operational conditions. The validity of the smartwatches was assessed using paired sample t-tests, mean absolute percentage errors (MAPE) and Bland–Altman plots.
Results - Both devices demonstrated a low average bias relative to the reference standard, with the Fenix 5 showing a small bias (0.20 ml·kg⁻¹·min⁻¹) and the Descent Mk2 exhibiting a higher positive bias (1.40 ml·kg⁻¹·min⁻¹), indicating a small tendency to overestimate VO₂ max. Both devices met the commonly accepted validity criterion of a MAPE ≤10% (Descent Mk2: 7.74%; Fenix 5: 9.86%).
Bland–Altman analysis revealed acceptable levels of agreement but substantial individual variability, particularly for the Fenix 5 (limits of agreement range: ~22.0 ml·kg⁻¹·min⁻¹) compared to the Descent Mk2 (~17.7 ml·kg⁻¹·min⁻¹). The narrower limits of agreement observed for the Descent Mk2 suggest a more consistent performance at the individual level. A key contextual factor was the difference in device familiarization: the Descent Mk2 had been used regularly by participants prior to testing, allowing its algorithms to calibrate more accurately to individual physiological patterns, whereas the Fenix 5 was worn only for two weeks before data collection, possibly contributing to greater variability.
Conclusion - From an operational perspective, both devices offer advantages in military and tactical environments, including low weight, portability, and minimal interference with mission tasks, enabling continuous physiological monitoring in the field. While neither device demonstrated sufficient precision to replace laboratory-based gas-exchange analysis for individual diagnostic purposes, both provide sufficiently accurate estimations for real-time tracking during training or deployment.