Physical activity is well documented as a beneficial intervention for the prevention and treatment of chronic disease, while physical inactivity is an independent risk factor for the development of lifestyle-related chronic diseases, such as cardiovascular disease [1,2]. To be considered “physically active” it is recommended that adults accumulate >150 minutes of moderate or 75 minutes of vigorous-intensity physical activity (MVPA) over the course of a week to provide them with substantial health benefits [3]. Given that the average adult expends approximately four metabolic equivalents walking at a moderate pace [4], they would therefore need to walk at this pace for 125 minutes/week, or 30 minutes on most days of the week, to meet the minimum recommendations. Although these physical activity guidelines are strongly recommended by governments across the world [5-7], many people do not achieve them [8-11]. However, physical activity interventions that have incorporated technology-based support have shown promise in developing compliance to physical activity guidelines [12,13], possibly through reinforcement to develop habit-forming behaviors [14].

Although valid and reliable tools for measuring physical activity are available for researchers, such as the ActiGraph [15], it is the quantified-self movement [16] that has led to the increasing popularity of consumer wearable technology, with estimates indicating sales of over 200 million by 2021, including the Apple Watch [17]. Having contemporary instruments to integrate into exercise prescription and physical activity promotion that fit into people’s lifestyles is imperative. Although evidence on the validity and reliability of modern wearable devices with integrated screens to measure exercise and physical activity is increasing, no study has examined the accuracy of the Apple Watch for measuring MVPA.

Apple is one of the world’s most valuable companies (by market value) [18] and has changed a number of industries through disruptive devices (eg, iPod, iPhone). The Apple Watch, which was released in 2015, has reportedly become the highest selling wearable/smartwatch to date, with more than 12 million units reported to have been sold [19,20]. The widespread use of the Apple Watch would therefore allow physical activity interventions to reach a large proportion of the population. Additionally, the Apple Watch provides potential for data to be collected from, and returned to, the individual, to provide immediate individualized feedback via the wearable’s screen to promote behavior change and be shared with others, such as a clinician. However, Apple provides little detail on how the Apple Watch measures “exercise.” It appears that the built-in accelerometer is mostly used to measure physical activity throughout the day [21,22], although heart rate measurement is used periodically when walking [22]. Daily physical activity data is displayed within the “Activity” app, with a visual representation of the accumulated duration of “exercise” displayed as a green ring on the screen of the watch. Given that Apple refers to this as the “exercise” ring and that 30 minutes is the daily goal [23], it is reasonable to suggest that Apple views this as representative of the daily goal of 30 minutes of MVPA for adults, as recommended by numerous guidelines [5-7]. Although Apple states that, “every full minute of movement that equals or exceeds the intensity of a brisk walk counts toward your daily exercise goal” [23], it is not clear how this measure of MVPA compares to the criterion measure of moderate-intensity exercise, oxygen consumption reserve (VO₂R: [VO₂max - VO₂rest] x exercise intensity + VO₂rest).

Given the scale of smartwatch use around the world and the increasing attention on personalized medicine, the validity of the Apple Watch for measuring moderate-intensity exercise is important to examine. Establishing the validity of the Apple Watch would ensure that individuals are able to measure their own MVPA accurately and that health professionals have confidence in the data that their clients are sharing with them. Establishing the intradevice and interdevice reliability of the Apple Watch is also important so that daily measures can be compared, and that user preference for wearing the Watch on the left or right wrist does not introduce bias. Therefore, the aim of this study was to examine the validity and reliability of the Apple Watch for quantifying moderate-intensity exercise compared with directly measured VO₂R.

The study used a repeated measures design with each participant completing two main trials. Prior to the main trials, maximal oxygen consumption (VO₂max) and resting oxygen consumption (VO₂rest) were measured in all participants. The study was approved by the Department of Sport, Health and Exercise Science Ethics Committee (approval number 1516076) at the University of Hull. Given the paucity of data on the Apple Watch at the time of study commencement, the sample size was estimated based on a previous study [24].

Participants were recruited from the University of Hull and local community via written promotional material or personal communication. Inclusion criteria stated that participants were aged between 18 and 50 years, and exclusion criteria were: (1) men and women classified as moderate or high-risk according to the American College of Sports Medicine (ACSM) risk classification criteria, (2) those unable to walk on a motorized treadmill, (3) current smokers, (4) BMI >30, and (5) those currently taking medication that alters the heart rate in response to exercise (eg, beta blockers). Inclusion and exclusion criteria were established with the aim of recruiting low-risk individuals, based on the ACSM risk classification criteria used at the commencement of the study.

Data were checked for normality using the Shapiro-Wilk test and graphical methods, and were found to be plausible. The VO₂R during Trial 2 was used to assess the validity of the Apple Watch for measuring moderate-intensity exercise, with VO₂R during both Trial 1 and 2 used to assess the interdevice and intradevice reliability. A one-sample t-test was used to test if the mean %VO₂R at the treadmill inflection speed for each Apple Watch was different from 40%, which is the lower limit of moderate-intensity exercise. A custom-designed Excel spreadsheet was used to examine differences between left and right Apple Watches for treadmill inflection speed and %VO₂R [27]. Pearson product-moment correlation was used to assess the association between VO₂max and the %VO₂R at the treadmill inflection speed. Based on the linear association between VO₂max and %VO₂R, linear regression was used to estimate the VO₂ max and treadmill speed required for the Apple Watch to accurately measure moderate-intensity exercise. Interdevice and intradevice reliability is reported as the standardized typical error and intraclass correlation. Standardized typical error was doubled prior to assessing its magnitude [28]. Standardized effect size is reported as Cohen’s d using the between-subject pooled SD as the denominator. The scale of magnitudes used to evaluate Cohen’s d was: 0-0.19 trivial; 0.2-0.59 small; 0.6-1.19 moderate; 1.2-1.99 large; >2.0 very large [28]. Uncertainty in the population estimates are reported as 95% CIs.

This is the first study to investigate the validity and reliability of the Apple Watch for measuring moderate-intensity exercise (the green “exercise” ring) compared to the criterion measure of VO₂R. The Apple Watch largely underestimated the walking speed required to elicit the lower bound of moderate-intensity exercise (40% VO₂R), which is an important part of individualized [3] and population-based [29] exercise prescription guidelines. The standardized typical error, a measure of the “typical” test-retest variability presented in units of SD [28], was small to moderate with no mean difference between the left and right watches for the mean treadmill inflection speed or %VO₂R for the lower limit of moderate-intensity exercise [3].

For less fit individuals (<35 mL/kg/minute), using the Apple Watch to monitor moderate-intensity exercise is more likely to have the expected and desired improvements to fitness or clinical outcomes than for fitter individuals. However, for those that are fitter (>35 mL/kg/minute) it appears from our data that individuals would not meet the expectations of the lower end (40% VO₂R) of the moderate-intensity range [3], with the Apple Watch underestimating the threshold for moderate-intensity exercise, and therefore overestimating the number of minutes of moderate-intensity exercise an individual had completed. Thompson et al [30] recently came to a similar conclusion, reporting that commercially-available wearable devices for self-monitoring of physical activity overestimate MVPA by a factor of 5-to-7-fold because they capture all physical activity, including normal moderate-to-vigorous lifestyle activities. However, other recent studies have reported that Fitbit devices (Fitbit One and Fitbit Flex) underestimate MVPA [31,32], although in these studies the wearable device was compared against another wearable device (ActiGraph), not direct laboratory-measured moderate-intensity exercise as we have done in our study. In our study only two participants achieved a %VO₂R within the moderate-intensity exercise zone (Figure 1). Apart from the impact this may have on the expected physiological adaptations and fitness, it also has potential implications for morbidity and mortality risk, as moderate fitness levels and habitual exercise have a protective effect for cardiovascular disease, stroke, type 2 diabetes, and all-cause mortality [33].

Measuring load in an individual is complicated with a variety of methods available to monitor internal and external load [34]. The Apple Watch does not substantially overestimate or underestimate heart rate [24] and has moderate interdevice variability of maximal heart rate when worn on each wrist [35]. Therefore, using a continuous combination of internal (heart rate) and external (accelerometer) training load measures [34,36] rather than using the accelerometer plus periodic use of heart rate as is currently used, may improve the accuracy of the green exercise ring and provide a greater personalization of the data available to an individual. Given the wide range of VO₂R responses observed at the treadmill speed that the Apple Watch determined to be “exercise” (<20% to >40%), and the variability between left and right Watches (Figure 1), it is clear that the Apple Watch needs to provide a more appropriate measure of the individualized response to exercise. This would not only enable more tailored feedback and exercise prescription, it would likely improve the physiological adaptations to any training and the associated cardiometabolic and musculoskeletal improvements for chronic disease prevention and treatment [33]. Better compliance to guidelines [13] and improved disease management and confidence [37] have been reported in healthy and chronic disease populations when technology-based support is incorporated. However, consumer and health professionals need to have confidence that any wearable device can both consistently and accurately measure the exercise intensity for individuals of all fitness levels.

The reliability data (Table 1) suggest that the reliability of VO₂R at the point where the Apple Watch determines that “exercise” has started (the treadmill inflection speed) is better within trials (interdevice) than between trials (intradevice). This finding would suggest that physical activity measurements are more reliable within the same training session or activity compared to between different training sessions or activities. The implication is that health professionals can be confident that the ability of the Apple Watch to measure physical activity (as used in the current study) is not adversely affected by the wrist (left or right) on which an individual wore the watch within a given training session or activity.

The main strength of our study is that this is the first investigation to examine the validity of the Apple Watch for measuring aspects of exercise related to the achievement of daily MVPA. The main limitation of this study is that we did not have direct access to the algorithms used by the Apple Watch for determining the exercise intensity at which the green “exercise” ring advances. Unfortunately, Apple does not publish these algorithms, probably due to commercial reasons. However, this is a limitation with most commercially-available wearable devices and is not restricted to the Apple Watch. A second limitation is that our data were derived from first generation Apple Watches. Although the method Apple uses to determine “exercise” using the Activity app has changed slightly with the periodic measurement of heart rate during walking [21,22], it is unclear how this would affect the validity of the Apple Watch for measuring MVPA. Further studies are required using the latest generation of Apple Watch for this to be determined. The exercise protocols used in our study were also constrained to linear walking on a treadmill. Movement patterns used by people outside of the laboratory while wearing an Apple Watch may result in different physiological responses and different determinations of “exercise” by the Apple Watch. Fourth, although our primary measure was VO₂R, there are a number of factors that can influence the oxygen cost of exercise, such as economy [38,39]. For example, economy has been reported to be affected by age, such that older adults have less economy when walking compared to younger people [40]. The mean age of our participants was 32 years (SD 10), which places our participants midway between the young and old adults participating in the study of Martin et al [40]. Although factors such as economy and age should be taken into account when interpreting our results, the use of %VO₂R as a relative measure of oxygen cost means that between-individual comparisons should still be meaningful.

The Series 0 Apple Watch underestimates the threshold for moderate-intensity exercise compared to the criterion measure of VO₂R, which would result in an overestimation of the amount of MVPA undertaken throughout the day. This effect is more pronounced in fitter individuals.