Comparative evaluation of heart rate-based monitors: Apple Watch vs Fitbit Charge HR

The purpose of this investigation was to examine the validity of energy expenditure (EE), steps, and heart rate measured with the Apple Watch 1 and Fitbit Charge HR. Thirty-nine healthy adults wore the two monitors while completing a semi-structured activity protocol consisting of 20 minutes of sedentary activity, 25 minutes of aerobic exercise, and 25 minutes of light intensity physical activity. Criterion measures were obtained from an Oxycon Mobile for EE, a pedometer for steps, and a Polar heart rate strap worn on the chest for heart rate. For estimating whole-trial EE, the mean absolute percent error (MAPE) from Fitbit Charge HR (32.9%) was more than twice that of Apple Watch 1 (15.2%). This trend was consistent for the individual conditions. Both monitors accurately assessed steps during aerobic activity (MAPE_Apple: 6.2%; MAPE_Fitbit: 9.4%) but overestimated steps in light physical activity. For heart rate, Fitbit Charge HR produced its smallest MAPE in sedentary behaviors (7.2%), followed by aerobic exercise (8.4%), and light activity (10.1%). The Apple Watch 1 had stronger validity than the Fitbit Charge HR for assessing overall EE and steps during aerobic exercise. The Fitbit Charge HR provided heart rate estimates that were statistically equivalent to Polar monitor.     

An adaptive filtering algorithm to estimate sprint velocity using a single inertial sensor

The assessment of sprint velocity is useful for evaluating performance and guiding training interventions. In this paper, we describe an adaptive filtering algorithm to estimate sprint velocity using a single, sacrum-worn magneto-inertial measurement unit. Estimated instantaneous velocity, average 10 m interval velocity, and peak velocity during 40 m sprints from the proposed method were compared to a reference method using photocell position-time data. Concurrent validity of the proposed method was assessed using mean absolute error and mean absolute percent error for all velocity estimates. The significance of the mean error was assessed using a factorial ANOVA for average interval velocity and a paired-samples t test for peak velocity. Reliability was assessed using Bland–Altman 95% limits of agreement for repeated measures. Average interval velocity was underestimated early in the sprint (??0.25 to ??0.05 m/s) and overestimated later (0.13 m/s) with mean absolute error between 0.20 m/s (3.95%) and 0.62 m/s (7.78%). The average mean absolute error was 0.45 m/s (7.02%) for instantaneous velocity and 0.63 m/s (7.84%) for peak velocity. The limits of agreement grew progressively wider at greater distances (??0.59 to 0.34 m/s for 0–10 m and ??1.32 to 1.59 m/s for 30–40 m). The estimation error from the proposed method is comparable to other wearable sensor-based methods and suggests its potential use to assess sprint performance.     

Assessment of step accuracy using the Consumer Technology Association standard

The purpose of this study was to compare the accuracy of commercially-available physical activity devices when walking and running at various treadmill speeds using CTA 2056: Physical Activity Monitoring for Fitness Wearables: Step Counting, standard by the Consumer Technology Association (CTA). Twenty participants (10 males and 10 females) completed self-paced walking and running protocols on the treadmill for five minutes each. Eight devices (Apple iWatch series 1, Fitbit Surge, Garmin 235, Moto 360, Polar A360, Suunto Spartan Sport, Suunto Spartan Trainer, and TomTom Spark 3) were tested two at a time, one per wrist. Manual step counts were obtained from video to serve as the benchmark. The mean absolute percent error (MAPE) was calculated during walking and running. During walking, three devices: Fitbit Surge (11.20%), Suunto Sport (22.93%), and TomTom (10.11%) and during running, one device, Polar (10.66%), exceeded the CTA suggestion of a MAPE < 10%. The Moto 360 had the lowest MAPE of all devices for both walking and running. The devices tested had higher step accuracy with running than walking, except for the Polar. Overall, the Apple iWatch series 1, Moto 360, Garmin, and Suunto Spartan Trainer met the CTA standard for both walking and running.     

Heart rate measures from the Apple Watch,Fitbit Charge HR 2, and electrocardiogram across different exercise intensities

This study compared heart rate (HR) measurements for the Fitbit Charge HR 2 (Fitbit) and the Apple Watch devices with HR measurements for electrocardiogram (ECG). Thirty young adults (15/15 females/males, age 23.5 ± 3.0 years) completed the Bruce Protocol. HR measurements were recorded from the ECG and both devices every minute. Average HR for each participant was calculated for very light, light, moderate, vigorous and very vigorous intensities based on ECG-measured HR. A concordance correlation coefficient (CCC) was calculated to examine the strength of the relationship between ECG measured HR and HR measured by each device. Relative error rates (RER) were also calculated to indicate the difference between each device and ECG. An equivalence test was conducted to examine the equivalence of HRs measured by devices and ECG. The Apple Watch showed lower RER (2.4–5.1%) compared with the Fitbit (3.9–13.5%) for all exercise intensities. For both devices, the strongest relationship with ECG-measured HR was found for very light PA with very high CCC (>.90) and equivalence. The strength of the relationship declined as exercise intensity increased for both devices. These findings indicate that the accuracy of real-time HR monitoring by the Apple Watch and Fitbit Charge HR2 is reduced as exercise intensity increases.     

Similar results for face mask versus mouthpiece during incremental exercise to exhaustion

Investigations in the 1990s evaluated the influence of breathing assemblies on respiratory variables at rest and during exercise; however, research on new models of breathing assemblies is lacking. This study compared metabolic gas analysis data from a mouthpiece with a noseclip (MOUTH) and a face mask (MASK). Volunteers (7 males, 7 females; 25.1 ± 2.7 years) completed two maximal treadmill tests within 1 week, one MOUTH and one MASK, in random order. The difference in maximal oxygen consumption (VO_2max) between MOUTH (52.7 ± 11.3 ml · kg^?1 · min^?1) and MASK (52.2 ± 11.7 ml · kg^?1 · min^?1) was not significant (P = 0.53). Likewise, the mean MOUTH–MASK differences in minute ventilation (V_E), fraction of expired oxygen (FEO₂) and carbon dioxide (FECO₂), respiration rate (RR), tidal volume (V_t), heart rate (HR), and rating of perceived exertion (RPE) at maximal and submaximal intensities were not significant (P > 0.05). Furthermore, there was no systematic bias in the error scores (r = ?0.13, P = 0.66), and 12 of the 14 participants had a VO_2max difference of ≤3 ml · kg^?1 · min^?1 between conditions. Finally, there was no clear participant preference for using the MOUTH or MASK. Selection of MOUTH or MASK will not affect the participant’s gas exchange or breathing patterns.     

Free-Living Comparison of Physical Activity and Sleep Data from Fitbit Activity Trackers Worn on the Dominant and Nondominant Wrists

Despite proprietary algorithms to account for differences, output from activity trackers worn on different wrists may not be comparable because individuals vary in their reliance on each hand during free-living activities.

Participants (n = 48) wore Fitbit Flex or Flex2 monitors on each wrist for three days. T tests, equivalence tests, and correlations were used to compare steps, Calories, distance, active minutes, and sleep duration recorded by dominant and nondominant wrist-worn monitors and effect sizes and mean absolute and percent difference were calculated.

The nondominant Flex2 monitor was not equivalent to the dominant wrist-worn monitor and recorded significantly more steps/day (absolute difference = 708), miles/day (0.3), and active minutes/day (7.9) than the dominant Flex2 monitor. For all variables, nondominant and dominant output was correlated (r>0.75).

Nondominant and dominant Flex2 monitors are significantly different, but there were small differences for Flex monitors. Research should investigate effects on behavior and replicate findings using other monitors.     

Validation of a single-stage submaximal treadmill walking test

Abstract

The single-stage treadmill walking test of Ebbeling et al. is commonly used to predict maximal oxygen consumption ([Vdot]O_2max) from a submaximal effort between 50% and 70% of the participant's age-predicted maximum heart rate. The purpose of this study was to determine if this submaximal test correctly predicts [Vdot]O_2max at the low (50% of maximum heart rate) and high (70% of maximum heart rate) ends of the specified heart rate range for males and females aged 18 – 55 years. Each of the 34 participants completed one low-intensity and one high-intensity trial. The two trials resulted in significantly different estimates of [Vdot]O_2max (low-intensity trial: mean 40.5 ml · kg^?1 · min^?1, s = 9.3; high-intensity trial: 47.5 ml · kg^?1 · min^?1, s = 8.8; P < 0.01). A subset of 22 participants concluded their second trial with a [Vdot]O_2max test (mean 47.9 ml · kg^?1 · min^?1, s = 8.9). The low-intensity trial underestimated (mean difference = ?3.5 ml · kg^?1 · min^?1; 95% CI = ?6.4 to ?0.6 ml · kg^?1 · min^?1; P = 0.02) and the high-intensity trial overestimated (mean difference = 3.5 ml · kg^?1 · min^?1; 95% CI = 1.1 to 6.0 ml · kg^?1 · min^?1; P = 0.01) the measured [Vdot]O_2max. The predictive validity of Ebbeling and colleagues' single-stage submaximal treadmill walking test is diminished when performed at the extremes of the specified heart rate range.     

Comparison of different theoretical models estimating peak power output and maximal oxygen uptake in trained and elite triathletes and endurance cyclists in the velodrome

Abstract

The primary objective of this study was to examine the relationship between heart rate intensity and pedometer step counts in adolescents. To determine cardiorespiratory fitness, 106 participants (47 boys, 59 girls, mean age 14.2 years, s = 0.8) completed the Queen's College Step Test and were classified as having low, moderate or high cardiorespiratory fitness. Adolescents also completed a 10-min treadmill trial while wearing a pedometer and heart rate monitor. The participants were instructed to maintain their heart rate between 65 and 75% of their maximum heart rate while running or walking on a treadmill. A heart rate of 65–75% maximum was associated with 146 steps per minute (s = 22) in boys and 137 steps per minute (s = 22) in girls. Results of analysis of variance indicated that there was a main effect for level of fitness (F _2,102 = 9.36, P < 0.001). The correlation between mean steps per minute and estimated maximum oxygen consumption was statistically significant (r = 0.44, P < 0.001). The results from this study suggest that a step rate of 130 steps per minute is equal to 65–75% maximum heart rate in low-fit adolescents and achieving 130 steps per minute could be used as an initial goal to improve fitness.     

The multi-stage fitness test as a predictor of endurance fitness in wheelchair athletes

Abstract

The aims of this study were two-fold: (1) to consider the criterion-related validity of the multi-stage fitness test (MSFT) by comparing the predicted maximal oxygen uptake ([Vdot]O_2max) and distance travelled with peak oxygen uptake ([Vdot]O_2peak) measured using a wheelchair ergometer (n = 24); and (2) to assess the reliability of the MSFT in a sub-sample of wheelchair athletes (n = 10) measured on two occasions. Twenty-four trained male wheelchair basketball players (mean age 29 years, s = 6) took part in the study. All participants performed a continuous incremental wheelchair ergometer test to volitional exhaustion to determine [Vdot]O_2peak, and the MSFT on an indoor wooden basketball court. Mean ergometer [Vdot]O_2peak was 2.66 litres · min^?1 (s = 0.49) and peak heart rate was 188 beats · min^?1 (s = 10). The group mean MSFT distance travelled was 2056 m (s = 272) and mean peak heart rate was 186 beats · min^?1 (s = 11). Low to moderate correlations (ρ = 0.39 to 0.58; 95% confidence interval [CI]: ?0.02 to 0.69 and 0.23 to 0.80) were found between distance travelled in the MSFT and different expressions of wheelchair ergometer [Vdot]O_2peak. There was a mean bias of ?1.9 beats · min^?1 (95% CI: ?5.9 to 2.0) and standard error of measurement of 6.6 beats · min^?1 (95% CI: 5.4 to 8.8) between the ergometer and MSFT peak heart rates. A similar comparison of ergometer and predicted MSFT [Vdot]O_2peak values revealed a large mean systematic bias of 15.3 ml · kg^?1 · min^?1 (95% CI: 13.2 to 17.4) and standard error of measurement of 3.5 ml · kg^?1 · min^?1 (95% CI: 2.8 to 4.6). Small standard errors of measurement for MSFT distance travelled (86 m; 95% CI: 59 to 157) and MSFT peak heart rate (2.4 beats · min^?1; 95% CI: 1.7 to 4.5) suggest that these variables can be measured reliably. The results suggest that the multi-stage fitness test provides reliable data with this population, but does not fully reflect the aerobic capacity of wheelchair athletes directly.     

Validity of Activity Trackers in Estimating Energy Expenditure During High-Intensity Functional Training

ABSTRACT

Purpose: The purpose of this study was to evaluate the agreement of five commercially available accelerometers in estimating energy expenditure while performing an acute bout of high-intensity functional training (HIFT). Methods: Participants (n = 47; average age: 28.5 ± 11.6 years) consisted of recreationally active, healthy adults. Each participant completed a session of HIFT: a 15-minute workout consisting of 12 repetitions each of air-squats, sit-ups, push-ups, lunges, pull-ups, steps-ups, and high-knees; performed circuit-style by completing as many rounds as possible. During this session, each participant wore the Cosmed K4b2 portable metabolic analyzer (PMA) and five different accelerometers (ActiGraph GT3X, Nike Fuelband, Fitbit One, Fitbit Charge HR, and Jawbone UP Move). Results: Four of the five activity trackers reported lower (p < .05) total EE values compared to the PMA during the acute bout of HIFT. The waist-mounted device (ActiGraph, 182.55 ± 37.93 kcal) was not significantly different from, and most closely estimated caloric expenditure compared to the PMA (144.99 ± 37.13 kcal) (p = .056). A repeated-measures ANOVA showed that all activity trackers were significantly different from the reference measure (PMA) (p < .05). Systematic relative agreement between the activity trackers was calculated, exhibiting a significant ICC = 0.426 (F [46,230] = 5.446 [p < .05]). Conclusion: The wrist- and hip-mounted activity trackers did not accurately assess energy expenditure during HIFT exercise. With the exception of the ActiGraph GT3X, the remaining four activity trackers showed inaccurate estimates of the amount of kilocalories expended during the HIFT exercise bout compared to the PMA.     

Validity of the Blast Athletic Performance Monitor for Assessing Vertical Jump Height in Female Volleyball Players

Introduction: Wearable activity monitors have been developed for jump height assessment, but the Blast Athletic Performance monitor has not yet been validated, and it remains unclear if the Blast can track changes across a sports season. Methods: Collegiate women’s volleyball players (n = 20) wore the Blast monitor (waistband) while performing standing vertical jumps (SVJs) and one-step vertical jumps (OSJs) weekly during and after a 9-week season. Jump heights from the Blast were compared to a Vertec (criterion). Results: Correlations of Blast and Vertec were moderately high (r = 0.67–0.69), but the Blast underestimated SVJ and OSJ (9.2–10.0 cm), with mean absolute percent errors 19.8–21.0%. A + 23% correction factor reduced errors to 10.5–11.3%. The Blast did not detect small decreases (2–4 cm) in criterion-measured jump height in the postseason. Conclusion: The Blast underestimated jump height and had limited ability to detect changes of up to 5.0 cm following a volleyball season. A relative correction lowered, but did not eliminate, measurement error.     

How valid are wearable physical activity trackers for measuring steps?

Wearable activity trackers have become popular for tracking individual’s daily physical activity, but little information is available to substantiate the validity of these devices in step counts. Thirty-five healthy individuals completed three conditions of activity tracker measurement: walking/jogging on a treadmill, walking over-ground on an indoor track, and a 24-hour free-living condition. Participants wore 10 activity trackers at the same time for both treadmill and over-ground protocol. Of these 10 activity trackers three were randomly given for 24-hour free-living condition. Correlations of steps measured to steps observed were r?=?0.84 and r?=?0.67 on a treadmill and over-ground protocol, respectively. The mean MAPE (mean absolute percentage error) score for all devices and speeds on a treadmill was 8.2% against manually counted steps. The MAPE value was higher for over-ground walking (9.9%) and even higher for the 24-hour free-living period (18.48%) on step counts. Equivalence testing for step count measurement resulted in a significant level within ±5% for the Fitbit Zip, Withings Pulse, and Jawbone UP24 and within ±10% for the Basis B1 band, Garmin VivoFit, and SenseWear Armband Mini. The results show that the Fitbit Zip and Withings Pulse provided the most accurate measures of step count under all three different conditions (i.e. treadmill, over-ground, and 24-hour condition), and considerable variability in accuracy across monitors and also by speeds and conditions.     

Reliability and validity of depth camera 3D scanning to determine thigh volume

Gross thigh volume is a key anthropometric variable to predict sport performance and health. Currently, it is either estimated by using the frustum method, which is prone to high inter-and intra-observer error, or using medical imaging, which is expensive and time consuming. Depth camera 3D-imaging systems offer a cheap alternative to measure thigh volume but no between-session reliability or comparison to medical imaging has been made. This experiment established between-session reliability and examined agreement with magnetic resonance imaging (MRI). Forty-eight male cyclists had their thigh volume measured by the depth camera system on two occasions to establish between-session reliability. A subset of 32 participants also had lower body MRIs, through which agreement between the depth camera system and MRI was established. The results showed low between-session variability (CV = 1.7%; Absolute Typical Error = 112 cm³) when measuring thigh volume using the depth camera system. The depth camera systematically measured gross thigh volume 32.6cm³ lower than MRI. These results suggest that depth camera 3D-imaging systems are reliable tools for measuring thigh volume and show good agreement with MRI scanners, providing a cheap and time-saving alternative to medical imaging analysis.     

Test–retest reliability of a battery of field-based health-related fitness measures for adolescents

Abstract

The main aim of this study was to determine the test–retest reliability of existing tests of health-related fitness. Participants (mean age 14.8 years, s = 0.4) were 42 boys and 26 girls who completed the study assessments on two occasions separated by one week. The following tests were conducted: bioelectrical impedance analysis (BIA) to calculate percent body fat, leg dynamometer, 90° push-up, 7-stage sit-up, and wall squat tests. Intra-class correlation (ICC), paired samples t-tests, and typical error expressed as a coefficient of variation were calculated. The mean percent body fat intra-class correlation coefficient was similar for boys (ICC = 0.95) and girls (ICC = 0.93), but the mean coefficient of variation was considerably higher for boys than girls (22.2% vs. 12.2%). The boys' coefficients of variation for the tests of muscular fitness ranged from 9.0% for the leg dynamometer test to 26.5% for the timed wall squat test. The girls' coefficients of variation ranged from 17.1% for the sit-up test to 21.4% for the push-up test. Although the BIA machine produced reliable estimates of percent body fat, the tests of muscular fitness resulted in high systematic error, suggesting that these measures may require an extensive familiarization phase before the results can be considered reliable.     

Converting between estimates of moderate-to-vigorous physical activity derived from raw accelerations measured at the wrist and from ActiGraph counts measured at the hip: the Rosetta Stone

The ability to compare published group-level estimates of objectively measured moderate-to-vigorous physical activity (MVPA) across studies continues to increase in difficulty. The objective of this study was to develop conversion equations and demonstrate their utility to compare estimates of MVPA derived from the wrist and hip. Three studies of youth (N = 232, 9-12yrs, 50% boys) concurrently wore a hip-worn ActiGraph and a wrist-worn GENEActiv for 7-days. ActiGraph hip count data were reduced using four established cutpoints. Wrist accelerations were reduced using the Hildebrand MVPA 200 mg threshold. Conversion equations were developed on a randomly selected subsample of 132 youth. Equations were cross-validated and absolute error, absolute percent error, and modified Bland-Altman plots were evaluated for conversion accuracy. Across equations R²_adj was 0.51–0.56 with individual-level absolute error in minutes ranging from 7 (wrist-to-hip Puyau) to 14.5 minutes (wrist-to-hip Freedson 3MET) and absolute percent differences ranging from 13.9%-24.5%. Group-level cross-validation to convert hip-to-wrist MVPA resulted in average absolute percent errors ranging from 3.1%-4.9%. Conversion of wrist-to-hip MVPA resulted in average absolute percent errors ranging from 3.0%-10.0%. We recommend the use of these equations to compare published estimates of MVPA between the wear-site cut-point combinations presented.     

The influence of ventilatory control on heart rate variability in children

The aim of this study was to determine the influence of breathing frequency and tidal volume on resting heart rate variability in children aged 9 years ( n = 29) and 16 years ( n = 19). Heart rate variability was measured in four conditions: (1) without the control of ventilation followed at random by (2) a fixed breathing frequency of 12 breaths· min -1 , (3) a breathing frequency of 12 breaths· min -1 but with a fixed tidal volume of 30% vital capacity and (4) a fixed breathing frequency of 6 breaths·min -1 and a tidal volume of 30% vital capacity. A total of 128 RR intervals (the time between two spikes in the heart rate) were detected and absolute high- and low-frequency spectral components were calculated using autoregressive modelling. The younger children were unable to control ventilation to achieve conditions 3 and 4; therefore, a 2 2 2 (group 2 condition) analysis of variance was used to analyse conditions 1 and 2. There were significant interactions between group and heart rate variability conditions for the low-frequency component and the ratio of low to high frequencies ( P ? 0.001). The main effect for condition showed that at 12 breaths· min -1 with no fixed tidal volume there was a significantly higher standard deviation of the RR interval, total power and high-frequency ( P ? 0.01) and lowfrequency spectral components ( P ? 0.05) than in the condition with no ventilatory control. Across the four breathing conditions for the older participants, the high-frequency spectral component was significantly higher in the condition at 6 breaths· min -1 with a fixed tidal volume than in that with no ventilatory control ( P ? 0.005); the ratio of high to low frequencies was significantly lower for the spontaneous condition than those performed at 12 breaths· min -1 ( P ? 0.001). The results provide evidence of the need for ventilatory control when assessing short-term resting heart rate variability in children.     

The influence of ventilatory control on heart rate variability in children

The aim of this study was to determine the influence of breathing frequency and tidal volume on resting heart rate variability in children aged 9 years (n = 29) and 16 years (n = 19). Heart rate variability was measured in four conditions: (1) without the control of ventilation followed at random by (2) a fixed breathing frequency of 12 breaths x min(-1), (3) a breathing frequency of 12 breaths x min(-1) but with a fixed tidal volume of 30% vital capacity and (4) a fixed breathing frequency of 6 breaths x min(-1) and a tidal volume of 30% vital capacity. A total of 128 RR intervals (the time between two spikes in the heart rate) were detected and absolute high- and low-frequency spectral components were calculated using autoregressive modelling. The younger children were unable to control ventilation to achieve conditions 3 and 4; therefore, a 2 x 2 (group x condition) analysis of variance was used to analyse conditions 1 and 2. There were significant interactions between group and heart rate variability conditions for the low-frequency component and the ratio of low to high frequencies (P < 0.001). The main effect for condition showed that at 12 breaths x min(-1) with no fixed tidal volume there was a significantly higher standard deviation of the RR interval, total power and high-frequency (P< 0.01) and low-frequency spectral components (P < 0.05) than in the condition with no ventilatory control. Across the four breathing conditions for the older participants, the high-frequency spectral component was significantly higher in the condition at 6 breaths x min(-1) with a fixed tidal volume than in that with no ventilatory control (P < 0.005); the ratio of high to low frequencies was significantly lower for the spontaneous condition than those performed at 12 breaths x min(-1) (P < 0.001). The results provide evidence of the need for ventilatory control when assessing short-term resting heart rate variability in children.     

Improving energy expenditure estimates from wearable devices: A machine learning approach

ABSTRACT

A means of quantifying continuous, free-living energy expenditure (EE) would advance the study of bioenergetics. The aim of this study was to apply a non-linear, machine learning algorithm (random forest) to predict minute level EE for a range of activities using acceleration, physiological signals (e.g., heart rate, body temperature, galvanic skin response), and participant characteristics (e.g., sex, age, height, weight, body composition) collected from wearable devices (Fitbit charge 2, Polar H7, SenseWear Armband Mini and Actigraph GT3-x) as potential inputs. By utilising a leave-one-out cross-validation approach in 59 subjects, we investigated the predictive accuracy in sedentary, ambulatory, household, and cycling activities compared to indirect calorimetry (Vyntus CPX). Over all activities, correlations of at least r = 0.85 were achieved by the models. Root mean squared error ranged from 1 to 1.37 METs and all overall models were statistically equivalent to the criterion measure. Significantly lower error was observed for Actigraph and Sensewear models, when compared to the manufacturer provided estimates of the Sensewear Armband (p < 0.05). A high degree of accuracy in EE estimation was achieved by applying non-linear models to wearable devices which may offer a means to capture the energy cost of free-living activities.     

Rain influences the physiological and metabolic responses to exercise in hot conditions

Outdoor exercise often proceeds in rainy conditions. However, the cooling effects of rain on human physiological responses have not been systematically studied in hot conditions. The present study determined physiological and metabolic responses using a climatic chamber that can precisely simulate hot, rainy conditions. Eleven healthy men ran on a treadmill at an intensity of 70% VO_2max for 30 min in the climatic chamber at an ambient temperature of 33°C in the presence (RAIN) or absence (CON) of 30 mm · h^?1 of precipitation and a headwind equal to the running velocity of 3.15 ± 0.19 m · s^?1. Oesophageal temperature, mean skin temperature, heart rate, rating of perceived exertion, blood parameters, volume of expired air and sweat loss were measured. Oesophageal and mean skin temperatures were significantly lower from 5 to 30 min, and heart rate was significantly lower from 20 to 30 min in RAIN than in CON (P < 0.05 for all). Plasma lactate and epinephrine concentrations (30 min) and sweat loss were significantly lower (P < 0.05) in RAIN compared with CON. Rain appears to influence physiological and metabolic responses to exercise in heat such that heat-induced strain might be reduced.