Assessment of heat strain considers an individual's tolerance and indicates the risk and physiological cost of working in hot environments. This study evaluated the discrimination ability of metrics of heat strain. The null hypotheses were that (1) the metrics individually could not discriminate between acceptable and unacceptable heat strain, (2) there were no significant differences among these metrics, and (3) there were no significant differences in the applicability of the metrics due to clothing or heat stress level. The experimental design was a case crossover. Clothing and heat stress level were potential confounders. Two clothing ensembles were work clothes and vapor-barrier coveralls with hood. Two heat stress levels for a moderate metabolic rate were 5°C-WBGT and 10°C-WBGT above the Threshold Limit Value adjusted for clothing. Eight male and four female acclimated individuals (age 18-36 years) participated. Four experimental trials were randomized in sequence. The transition point, when a participant's status changed from control (acceptable heat strain) to case (unacceptable), was the first occurrence of rectal temperature equal to or greater than 38.5°C, heart rate equal to or greater than 90% of maximum, or volitional fatigue. The metrics were rectal, ear canal, oral, and disk temperatures, heart rate including moving time averages of 5, 10, 20, 30 and 45 minutes, recovery heart rate, and physiological strain index. The data at the transition point were the case data; the data 10 minutes prior to that point were the control data. Analyses used primarily receiver operating characteristic (ROC) curves, which indicated the ability to distinguish acceptable from unacceptable heat strain. Further analyses included factorial analysis of variance and exact conditional logistic regression. Based on the ROC curve analyses, the physiological metrics can distinguish between acceptable and unacceptable heat strain with average area under the curves between 0.529 and 0.861. While there were no differences among the metrics based on the 95% confidence intervals of the areas under the curve, the results were compromised by low power. Based on ANOVA and logistic regression, clothing did not influence the metrics. There were insufficient data to evaluate the role of heat stress level.