Author(s): DE BOT, I., GIELEN, J., MEEUSEN, R., ROELANDS, B., AERTS, J.M., Institution: KU LEUVEN, Country: BELGIUM, Abstract-ID: 615

INTRODUCTION:
As athletes are increasingly exposed to hot conditions in training and competition due to rising global temperatures, reliable monitoring of physiological responses to thermal load becomes essential. Wearables offer practical opportunities to track core temperature (Tc), skin temperature (Ts), heart rate (HR), and related physiological signals. However, despite rapid technological advances, uncertainty remains regarding how accurately these devices quantify heat strain. Therefore, our aim was to synthesize evidence from studies evaluating wearables and computational models for assessing heat strain and heat related risks in various settings, including sport and exercise, or studies with direct applicability to sport due to their focus on thermoregulatory responses during physical activity.
METHODS:
This systematic review was conducted in line with the PRISMA framework. PubMed, ScienceDirect and Web of Science were searched for relevant studies in November 2025. Studies were eligible if they examined wearables to monitor physiological parameters under conditions associated with heat stress or heat-exacerbating factors.
RESULTS:
In total, 48 studies met all inclusion criteria and were grouped into four clusters:
(1) Wearable validation (n = 21): several sport-applicable devices (e.g., CORE, Zephyr) showed variable agreement with Tc reference measures (bias = -0.47 to 0.51 °C), with accuracy influenced by, inter alia, environmental conditions, sensor placement, and activity intensity. Bias often increased at higher temperatures, indicating reduced reliability during hyperthermia.
(2) Algorithmic Tc estimation (n = 9): compared with traditional linear or mechanistic models, Kalman filter and deep learning approaches, particularly those using multimodal inputs and phase specific update procedures, achieved superior accuracy (bias = -0.06 to 0.30 °C).
(3) Real time heat strain monitoring (n = 9): wearables reliably tracked key indicators of heat strain. While much evidence stems from occupational settings, the findings are transferable to sport and exercise, where athletes may experience similar cardiovascular and thermal strain. Notably, perceived strain often exceeded physiological measures, reinforcing the value of objective, individualized feedback.
(4) Predictive analytics for heat stress risk and heat strain estimation (n = 9): machine learning models that integrate wearable-derived physiological signals, environmental and contextual data improved early detection and prediction of excessive heat strain and heat related risks.
CONCLUSION:
Our review shows that wearables and predictive modelling show potential for individualized, real time assessment of heat strain and heat related risks. Yet, rigorous validation, physiologically grounded modelling, and broader population testing are essential for improving reliability, generalizability and real world applicability, especially within sport, where the applicability of existing findings and sport specific validation remain limited.