JMIR Publications

ABSTRACT

Background:

Cardiorespiratory fitness (CRF) is an independent risk factor for cardiovascular morbidity and mortality. Indeed, adding CRF to conventional risk factors (e.g., smoking, hypertension, impaired glucose metabolism, dyslipidemia) improves the prediction of individual’s risk for adverse health outcomes such as for those related to cardiovascular disease. It is consequently recommended to determine CRF as part of individualized risk prediction. However, CRF is not determined routinely in everyday clinical practice. Wearable technologies provide one potential strategy to estimate CRF on a daily basis, and such technologies, providing CRF estimates based on heart rate and body acceleration, have been developed. However, the validity of such technologies to estimate individual CRF in clinically relevant populations is poorly known.

Objective:

The objective of this study was to evaluate the validity of one wearable technology, providing estimated CRF based on heart rate and body acceleration, in working-age adults with cardiovascular risk factors.

Methods:

Seventy-four adults (age range 35-64 years, 76% female, body mass index (BMI) 28.7 kg/m2 (standard deviation 4.6)) with frequent cardiovascular risk factors (e.g., 86% hypertension, 24% prediabetes, 19% type 2 diabetes, 69% metabolic syndrome) performed a 30-min self-paced walk on an indoor track and a cardiopulmonary exercise test (CPET) on a treadmill. CRF, quantified as peak O2 uptake (V̇O2peak), was both estimated (self-paced walk: a wearable single-lead electrocardiogram device worn to record continuous beat-to-beat R-R intervals and triaxial body acceleration) and measured (CPET: ventilatory gas analysis). The accuracy of estimated CRF was evaluated against measured CRF.

Results:

Measured CRF averaged 30.6 ml/kg/min (range 20.1-49.6). In all participants, mean difference between estimated and measured CRF was -0.1 ml/kg/min (P=.90), mean absolute error (MAE) was 3.1 ml/kg/min (95% confidence interval (CI) 2.6, 3.7), mean absolute percentage error (MAPE) was 10.4 % (95% CI 8.5, 12.5), and intraclass correlation coefficient was 0.88 (95% CI 0.80, 0.92). Similar accuracy was observed in various subgroups (sexes, age and BMI categories, hypertension, prediabetes, metabolic syndrome). Instead, MAE was 4.2 ml/kg/min (95% CI 2.6, 6.1) and MAPE was 16.5 % (8.6, 24.4) in the subgroup of 14 patients with type 2 diabetes.

Conclusions:

The error of the CRF estimate, provided by the wearable technology, was likely below or at least very close to the clinically significant level of 3.5 ml/kg/min in working-age adults with cardiovascular risk factors but not in the relatively small subgroup of type 2 diabetes patients. From a large-scale clinical perspective, the findings suggest wearable technologies carry potential to estimate individual CRF with acceptable accuracy in clinically relevant populations.