Cross-validation Of Whole Body Sweat Sodium Prediction Equations

Abstract

We have previously published regression equations to estimate whole body (WB) sweat sodium concentration ([Na]) from regional (REG) measures; however, a cross-validation is needed to corroborate the applicability of these prediction equations between studies. PURPOSE: To determine the validity of published regression equations (Baker et al. 2018) in predicting WB sweat [Na] from REG measures when applied to a new data set. METHODS: Forty-nine recreational athletes (34 men, 15 women; 75±12 kg) cycled for 90 min while WB sweat [Na] was measured using the washdown technique. Exercise intensity (82% HRmax) and environmental conditions (32°C, 39% rh, 2.4 m/s air flow) were similar to the 2018 study in which the prediction equations were developed. REG sweat [Na] was measured from seven regions using absorbent patches (3M Tegaderm+Pad, 10 cm2). Regression equations from Baker et al. 2018 were applied to REG sweat [Na] to determine predicted WB sweat [Na]. Bland-Altman analysis of mean bias (raw and predicted minus measured) and 95% limits of agreement (LOA) were used to compare raw (uncorrected) REG sweat [Na] and predicted WB sweat [Na] to measured WB sweat [Na]. RESULTS: Mean±SD WB sweating rate was 0.94±0.32 L/h and measured WB sweat [Na] was 41±16 mmol/L. Mean bias (±95% LOA) between raw REG sweat [Na] and measured WB sweat [Na] was 10(±20), 0(±19), 9(±20), 22(±25), 23(±24), 0(±15), -4(±18) mmol/L for the dorsal forearm, ventral forearm, upper arm, chest, upper back, thigh, and calf, respectively. The mean bias (±95% LOA) between predicted WB sweat [Na] and measured WB sweat [Na] was 3(±14), 4(±12), 0(±14), 2(±17), -2(±16), 5(±13), 4(±15) mmol/L for the dorsal forearm, ventral forearm, upper arm, chest, upper back, thigh, and calf, respectively. CONCLUSIONS: The use of regression equations enables prediction of WB sweat [Na] within a mean bias of 0-5 mmol/L and within a 95% LOA of ±12-17 mmol/L across all sites. By contrast, the use of raw REG [Na] increases the mean bias to 9-23 mmol/L for the dorsal forearm, upper arm, chest, and upper back, and increases the 95% LOA to ±15-25 mmol/L across all sites. Regression equations improve the accuracy of estimating WB sweat [Na] from REG measures and are therefore recommended for use in Na balance studies and field tests to determine individualized sweat electrolyte losses.