Minimally invasive prediction of blood lactate during incremental exercise via heart rate, core body temperature, and sweat-derived indices.

Blood lactate concentration (BLa) is a key marker of metabolic stress, but invasive sampling limits real-time monitoring. We developed a minimally invasive model to estimate BLa during incremental exercise using heart rate (HR), core body temperature (CBT), and sweat-derived indices. Thirty-one healthy adult males performed a graded treadmill test. HR and CBT were monitored continuously. Sweat was sampled from the forehead, chest, and back to quantify sweat lactate concentration ([La⁻]sw) and lactate excretion rate (LER = [La⁻]sw × sweat rate). Linear mixed-effects models (LMMs) were fitted with log-transformed BLa (Log[BLa]) and participant-level random effects. BLa increased with exercise intensity (p < 0.001), accompanied by increases in HR, CBT and LER (both p < 0.001). LMMs combining HR, CBT, and sweat indices showed strong performance for Log[BLa]. The best model (HR + CBT+forehead LER) achieved conditional R²=0.939 and RMSE = 0.229 (log units), and forehead-based models outperformed chest and back. Combined cardiovascular, thermoregulatory, and sweat-derived measures enable accurate, minimally invasive estimation of BLa during graded exercise, supporting wearable-based metabolic monitoring and individualized exercise prescription.