Access to this full-text is provided by Springer Nature.
Content available from Sports Medicine
This content is subject to copyright. Terms and conditions apply.
Vol.:(0123456789)
Sports Medicine (2022) 52:1577–1597
https://doi.org/10.1007/s40279-021-01639-y
SYSTEMATIC REVIEW
Validity ofEstimating theMaximal Oxygen Consumption byConsumer
Wearables: ASystematic Review withMeta‑analysis andExpert
Statement oftheINTERLIVE Network
PabloMolina‑Garcia1,2 · HannahL.Notbohm3· MoritzSchumann3,4· RobArgent5,6,7·
MeganHetherington‑Rauth8· JulieStang9· WilhelmBloch3· SulinCheng3,4· UlfEkelund9· LuisB.Sardinha8·
BrianCauleld5,6· JanChristianBrønd10· AndersGrøntved10· FranciscoB.Ortega1,11,12
Accepted: 20 December 2021 / Published online: 24 January 2022
© The Author(s) 2022
Abstract
Background Technological advances have recently made possible the estimation of maximal oxygen consumption (VO2max)
by consumer wearables. However, the validity of such estimations has not been systematically summarized using meta-
analytic methods and there are no standards guiding the validation protocols.
Objective The aim was to (1) quantitatively summarize previous studies investigating the validity of the VO2max estimated
by consumer wearables and (2) provide best-practice recommendations for future validation studies.
Methods First, we conducted a systematic review and meta-analysis of studies validating the estimation of VO2max by
wearables. Second, based on the state of knowledge (derived from the systematic review) combined with the expert discus-
sion between the members of the Towards Intelligent Health and Well-Being Network of Physical Activity Assessment
(INTERLIVE) consortium, we provided a set of best-practice recommendations for validation protocols.
Results Fourteen validation studies were included in the systematic review and meta-analysis. Meta-analysis results
revealed that wearables using resting condition information in their algorithms significantly overestimated VO2max (bias
2.17ml·kg−1·min−1; limits of agreement − 13.07 to 17.41ml·kg−1·min−1), while devices using exercise-based information
in their algorithms showed a lower systematic and random error (bias − 0.09ml·kg−1·min−1; limits of agreement − 9.92 to
9.74ml·kg−1·min−1). The INTERLIVE consortium proposed six key domains to be considered for validating wearable devices
estimating VO2max, concerning the following: the target population, reference standard, index measure, testing conditions,
data processing, and statistical analysis.
Conclusions Our meta-analysis suggests that the estimations of VO2max by wearables that use exercise-based algorithms
provide higher accuracy than those based on resting conditions. The exercise-based estimation seems to be optimal for
measuring VO2max at the population level, yet the estimation error at the individual level is large, and, therefore, for sport/
clinical purposes these methods still need improvement. The INTERLIVE network hereby provides best-practice recom-
mendations to be used in future protocols to move towards a more accurate, transparent and comparable validation of VO2max
derived from wearables.
PROSPERO ID CRD42021246192.
* Pablo Molina-Garcia
pablomolinag5@gmail.com
* Francisco B. Ortega
ortegaf@ugr.es
Extended author information available on the last page of the article
1 Introduction
The use and development of wearable technology monitor-
ing fitness and activity have grown exponentially over the
last few years. In 2020, 396 million wearable units were
shipped worldwide, and it is forecasted that this will increase
up to 631.7 million units by 2024 [1]. Wearable devices give
users the opportunity to monitor health-related metrics, such
as daily steps, heart rate (HR), energy expenditure, or cardi-
orespiratory fitness, therefore, promoting physical activity
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
1578 P.Molina-Garcia et al.
Key Points
Wearables using exercise-based algorithms provide
higher accuracy in the estimation of maximal oxygen
consumption (VO2max) than those based on resting condi-
tions.
Wearables using exercise-based estimation seem to be
optimal for measuring VO2max at the population level,
yet the estimation error at the individual level still needs
further improvement.
In this article, the Towards Intelligent Health and
Well-Being Network of Physical Activity Assessment
(INTERLIVE) network provides best-practice recom-
mendations to be used in future protocols to move
towards a more accurate, transparent and comparable
validation of VO2max derived from wearables.
reviews have already assessed how well wearable devices
estimate most of the health measures such as step count [12,
13], HR [14, 15], and energy expenditure [14, 16]; how-
ever, to the best of our knowledge, no systematic review
or meta-analysis focusing on the validity of the estimated
VO2max is available. Furthermore, the current science behind
the validation protocols of wearable devices suffers major
limitations, mainly due to a lack of consensus and guidelines
ensuring good practices [17, 18]. This is precisely one of
the main goals of the Towards Intelligent Health and Well-
Being Network of Physical Activity Assessment (INTER-
LIVE) consortium, which is to develop best-practice pro-
tocols for the validation of consumer wearable fitness and
activity measures. The INTERLIVE consortium has already
published guidelines adapted to the nature of specific fitness/
physical activity measures such as step count [19] and HR
[20]. However, to date there are no specific standards guid-
ing both manufacturers and the scientific community in the
validation of estimating VO2max by consumer wearables.
Therefore, in this article, INTERLIVE had two main
objectives: (1) to systematically summarize previous stud-
ies investigating the validity of VO2max as estimated by con-
sumer wearable devices based on a meta-analysis, and (2)
to provide best-practice validation recommendations based
on the systematic review of the literature together with an
evidence-informed INTERLIVE consortium discussion.
2 Methods: Expert Statement Process
andMeta‑Analysis
2.1 The INTERLIVE Network
INTERLIVE (https:// www. inter live. org/) is a consortium
composed of six universities—University of Lisbon (Por-
tugal), German Sport University (Germany), University of
Southern Denmark (Denmark), Norwegian School of Sport
Sciences (Norway), University College Dublin (Ireland),
and University of Granada (Spain)—and one technology
company, Huawei Technologies (Finland). The consortium
was founded in 2019 and strives towards developing best-
practice protocols for evaluating the validity of consumer
wearables with regard to the measurement of exercise/activ-
ity metrics. Moreover, INTERLIVE aims to increase aware-
ness of the advantages and limitations of different validation
methods and to introduce novel health and performance-
related metrics, fostering a widespread use of physical activ-
ity indicators.
2.2 Expert Validation Process
The consortium followed the same process as was used pre-
viously [19, 20]. First, we conducted a systematic review
and optimizing health and sports performance [2, 3]. Fur-
thermore, the omnipresence of wearables enhances digital
phenotyping at a population level, which offers valuable
information about physical activity and fitness levels from
around the world that can be used to guide global health
promotion actions [2, 4].
The most accepted measure of cardiorespiratory fitness
is maximal oxygen consumption (VO2max), which has been
shown to be a powerful marker of health and has recently
been proposed as a clinical vital sign by the American Heart
Association [5]. Furthermore, VO2max is widely known as a
key indicator of endurance performance and, therefore, its
measurement is of vital importance for sports performance
in general [6]. The current guidelines for accurate testing
of VO2max require measurement of gas exchange by indirect
calorimetry usually in a laboratory during an exercise test
to exhaustion [7]. These tests require expensive equipment
(e.g., gas analyzer) and trained technicians to collect and
interpret the data, which makes VO2max assessments less
feasible for risk prediction in clinical practice and unaf-
fordable for most recreational athletes and for the general
population. Indirect estimation of VO2max by submaximal
field tests overcomes some of these disadvantages and offers
acceptable estimations of VO2max [8, 9]. However, the above-
mentioned digital era of consumer wearable devices opens
new horizons for fitness monitoring without the need for
laboratory or field testing.
In view of the enormous potential of these devices,
wearable companies are making significant investments in
research and development to provide valid fitness and activ-
ity measures, such as VO2max [10, 11]. Previous systematic
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
1579
Validity of VO2max Estimated by Wearables
of the scientific literature on the studies validating VO2max
estimated by consumer wearables against a reference stand-
ard (criterion measure). Second, the information obtained
from the systematic review, together with previous related
statements [17–21], was critically discussed within the
consortium to provide guidelines and recommendations on
how to conduct optimal validation protocols. Third, a set of
key domains for best-practice recommendations was pro-
posed based on the evidence-informed expert opinion of the
INTERLIVE members.
2.3 Systematic Review andMeta‑Analysis Process
This systematic review was guided by the Preferred
Reporting Items for Systematic Reviews and Meta-Anal-
yses diagnostic test accuracy guideline. The protocol was
registered in advance in the PROSPERO database (ID:
CRD42021246192).
2.3.1 Data Sources andSearch Strategy
PubMed, Web of Sciences, and Scopus databases were
searched dating up to January 14, 2021. Members from the
INTERLIVE network defined the search strategy, which
can be found for replication in Supplementary Material 1
(see the electronic supplementary material). Additionally, a
hand-search using the same search strategy was performed
in Google Scholar to identify additional studies.
2.3.1.1 Inclusion and Exclusion Criteria We considered
studies meeting the following criteria: (1) any kind of pop-
ulation, (2) VO2max estimated through consumer wearable
devices and measured with the reference standard (a graded
exercise test to exhaustion with direct or indirect [gas anal-
ysis] calorimetry using a mode of test that involves large
muscle groups), and (3) criterion validity studies.
We excluded studies following these criteria: (1) non-
consumer wearable devices (e.g., research-based accelerom-
eters), (2) not original articles (e.g., reviews or editorials)
and grey literature (e.g., meeting abstracts), and (3) articles
validating new algorithms in the estimation of VO2max that
are not yet incorporated in any commercial brand.
2.3.2 Study Selection
Two authors (PM-G and HLN) independently performed
both the title, abstract, and full-text screening of potential
articles and any discrepancy was solved in a consensus meet-
ing with a third author (MS). This systematic review process
was performed using the Covidence software (www. covid
ence. org; Veritas Health Innovation).
2.3.3 Data Extraction
For each included article we extracted the following infor-
mation: (1) author’s name and publication year, (2) target
population (e.g., healthy adults), sample size, and age range,
(3) protocol used for the VO2max assessment via reference
standard (e.g., indirect calorimetry), (4) gas analyzer brand
used, (5) wearable device used, (6) protocol followed for the
estimation of VO2max via wearable devices, and (7) statistical
analysis used to test the validity of wearable VO2max against
the reference standard. Two independent authors (PM-G and
HLN) performed the data extraction, and any discrepancies
were discussed until consensus was reached.
2.3.4 Risk ofBias
The Consensus-based Standards for the selection of health
Measurement Instruments (COSMIN) checklist was adapted
and used to assess the risk of bias of included studies. The
COSMIN checklist contains standards for evaluating the
methodological quality of studies validating health meas-
urement instruments [22], and it encompasses four domains:
(1) participants included, (2) index measure (i.e., wearable
device), (3) reference standard (i.e., indirect calorimetry),
and (4) statistical analysis. Each domain contains several
items with three possible answers (“yes,” “unclear,” and
“no”) according to the fulfillment of the criterion and, there-
fore, the presence or absence of bias (Supplementary Mate-
rial 2; see the electronic supplementary material). According
to the Risk of Bias 2 (RoB 2) criteria proposed by Cochrane
[23], an article having at least one “no” or more than two
“unclear” items was categorized as having “high risk” of
bias; having one “unclear” item was categorized as “some
concerns” in the risk of bias; and having all items answered
as “yes” was categorized as “low risk” of bias. Two inde-
pendent researchers (PM-G and AG) accomplished this pro-
cess, and disagreements were discussed to reach a consensus
including a third author (FBO).
2.3.5 Meta‑Analysis
We identified two main methodologies to estimate VO2max
through wearable devices: (1) the resting conditions that
evaluate users lying in a supine position and/or standing
still, and (2) exercise-based methodologies that evaluate
users while performing physical activity. Therefore, we per-
formed and reported the meta-analysis separately for these
two methods—the resting and exercise tests. The bias of the
estimation of VO2max by the wearables (i.e., the mean differ-
ence between the wearable and the reference standard) and
the standard errors of this bias in all included studies were
used to calculate the pooled bias and its 95% confidence
interval (CI) for both the resting and exercise test. A negative
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
1580 P.Molina-Garcia et al.
bias represents an underestimation of the wearable VO2max
relative to the reference VO2max, while a positive value rep-
resents an overestimation. The Higgins I2 statistic and P
value were used to test the heterogeneity of included studies,
which were classified as not important (0–40%), moderate
(30–50%), substantial (50–75%), or considerable (75–100%)
[24]. Due to the presence of considerable heterogeneity in
both meta-analyses (Higgins I2 = 77% and 88% in resting and
exercise test, respectively), we used a random-effects model
of the inverse variance method. Klepin etal. [25] averaged
the gas exchange data every 15 and 60s, and we selected the
15s time averaging according to previous recommendations
[26]. Two studies examined the wearable validity separately
in men and women [27, 28], and we maintained this divi-
sion when including the data in the meta-analysis. There
were five studies [29–31] that did not report the bias to test
the validity or reported it in plots. Therefore, validity was
estimated from correlation coefficients between the wear-
able and reference VO2max, as suggested elsewhere [32], or
extracted from plots through the WebplotDigitizer software
(Ankit Rohatgi, website: https:// autom eris. io/ WebPl otDig
itizer/), which has demonstrated an excellent validity and
reliability in extracting graphed data [33].
The framework for the meta-analysis of Bland–Altman
studies proposed by Tipton and Shuster [34] was used to
obtain a pooled limit of agreement in both the resting and
exercise test, which was calculated with the following for-
mula: δ ± 2 √ σ2 + τ2, where δ is the average bias across
studies, σ2 is the average within-study variation in differ-
ences, and τ2 is the variation in bias across studies [34]. The
weighted least-squares models from the abovementioned
random-effect meta-analysis were used to estimate δ and
σ2, while the DerSimonian and Laird procedure was used to
estimate τ2 [35]. The R code provided in the study of Tipton
and Shuster [34] was used to conduct all these analyses with
the RStudio statistical program.
Three sensitivity analyses were performed: (1) to test
the robustness of the results, (2) to evaluate the presence of
publication bias, and (3) to divide the meta-analyses results
into those studies using photoplethysmography (PPG) tech-
nology to assess HR versus those using chest straps. For the
robustness analysis, studies were removed one at a time and
we tested whether the overall effect size (i.e., z score and P
value) was significantly modified in magnitude or direction.
The publication bias was assessed by a funnel plot and the
Egger regression asymmetry test, considering the level of
significance < 0.100 [36]. The meta-analysis was repeated
in the two following conditions: (1) splitting the results into
studies using PPG and chest straps to measure HR and (2)
including studies from the last 3 years. Thus, we tested the
impact of the different types of HR recordings (PPG vs.
chest straps) and of old articles testing obsolete devices on
the error estimates.
The meta-analysis was performed using the Review
Manager Version 5.3 (The Nordic Cochrane Center, The
Cochrane Collaboration, 2014, Copenhagen, Denmark),
and the limit of agreement meta-analyses were performed
using the RStudio statistical program (version 1.4.1106, R
Core Team 2020; R Foundation for Statistical Computing,
Vienna, Austria; https:// www.R- proje ct. org/).
3 Results
3.1 Summary oftheIncluded Studies
intheSystematic Review
The flow chart (Fig.1) shows that among the 1224 non-
duplicated studies initially included, 1189 were excluded
after the first screening of title and abstract and another 27
were further excluded after the full-text screening. Con-
sequently, 14 articles meeting the inclusion criteria were
included in the systematic review and the meta-analysis;
eight and eight studies reporting on the validity of an exer-
cise-based and resting state-based methodology, respec-
tively, were included. Table1 summarizes the main informa-
tion extracted from the 14 included studies, including a total
of 403 participants. The risk of bias assessment of included
studies is reported in Fig.2 and Supplementary Material 3
(see the electronic supplementary material). The overall risk
of bias assessed across all domains was deemed to be “some
concerns” for three (21%) and “high” for 11 (79%) of the 14
studies included.
3.2 Validity oftheVO2max Estimated byWearables:
Meta‑Analysis
The forest plots with the pooled bias between the reference
VO2max and the wearable estimation are presented in Fig.3
for both the wearables using the resting methodology and the
exercise test. Wearables using the resting test significantly
overestimated VO2max (bias = 2.17ml·kg−1·min−1; 95% CI
0.28–4.07; P = 0.020) in comparison to the reference stand-
ard. On the other hand, wearables estimating VO2max through
exercise tests showed a bias close to nil compared to the
reference standard (bias = − 0.09ml·kg−1·min−1; 95% CI
− 1.66 to 1.48; P = 0.910). Sensitivity analysis showed a lack
of robustness in the resting test meta-analysis since results
were significantly modified when removing five individual
studies [27, 28, 37–39], while the exercise test meta-analysis
indeed demonstrated robustness (Supplementary Material
4; see the electronic supplementary material). After a vis-
ual observation of the funnel plot and confirming with the
Egger’s tests, we did not find evidence of publication bias
either in the resting test or exercise test studies (Supple-
mentary Material 5). Studies using PPG technology in the
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
1581
Validity of VO2max Estimated by Wearables
HR recording had significantly greater bias than those using
chest strap in resting conditions, while the difference was not
statistically significant in the exercise testing methodology
(Supplementary Material 6 and 7). Finally, we excluded five
articles from more than 3years ago in the resting conditions
and we observed a significant reduction in the estimation
errors (bias = 1.66ml·kg−1·min−1; 95% CI − 0.58 to 3.90;
P = 0.150).
The Bland–Altman plot (Fig.4) presents the pooled
bias and its limits of agreement for both the resting and
exercise methodologies. The limits of agreements in the
resting test spanned from − 13.07 to 17.41ml·kg−1·min−1
(i.e., ± │15.24│; 95% CI − 22.18 to 26.53), while limits
were narrower in the exercise tests, spanning from − 9.92
to 9.74ml·kg−1·min−1 (i.e., ± │9.83│; 95% CI − 16.79 to
16.61). Therefore, the difference in limits of agreement was
smaller by 5.4ml·kg−1·min−1 in exercise tests compared to
the resting conditions. The limits of agreement in the differ-
ent studies using the resting conditions ranged from ± 17.75
[40] to ± 38.97 ml·kg−1·min−1 [41], while it spanned
from ± 11.18 [42] to ± 23.53ml·kg−1·min−1 [25] in the exer-
cise tests. Lastly, studies using PPG technology in the HR
recording had a greater span of the limits of agreement in
comparison with those using chest strap in the exercise tests
Fig. 1 Flowchart of the system-
atic review process
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
1582 P.Molina-Garcia et al.
Table 1 Characteristics of included studies (N = 14)
References Participants Age (years) Wearable device.
HR assessment Setup information VO2max estimation Reference standard VO2max protocol Statistical analysis
Anderson etal.
2019 [29]25 recreational
runners, men (17)
and women (8)
39.4 ± 10.8 Garmin Fenix 5X.
Wrist-measured
HR (PPG)
Age, sex, height,
and weight Exercise test:
walking or
jogging warm-
up + 10-min run
at their high-
est perceived
pace + 5-min cool
down walking
Indirect calorime-
try: ParvoMedics
TrueOne 2400
Treadmill: Bruce
running protocol
(speed and incli-
nation increase
each 3min)
T test and Pear-
son’s r
Carrier etal. 2020
[44]17 recreational
runners, men (8)
and women (9)
24.8 ± 4.3 Garmin Fenix
3 + chest HR strap HRmax and unspeci-
fied info Exercise test:
15-min outdoor
run above 70%
HRmax
Indirect calorime-
try: ParvoMedics Treadmill: modi-
fied Costill-Fox
running protocol
(speed increase
first and 2% incli-
nation increase
second each
2min)
T test, MAPE, Pear-
son correlation and
Bland–Altman
Cooper and Shafer
2019 [47]19 healthy, men (9)
and women (10) 21.9 ± 4.2 Polar A300 + chest
HR strap Age, sex, height,
and weight Resting HR: 5min
supine position Indirect calorim-
etry: Cosmed
Fitmate Pro
Treadmill: Bruce
running protocol
(speed and incli-
nation increase
each 3min)
Pearson’s r and
ANOVA
Crouter etal. 2004
[27]20 active men (10)
and women (10) Men: 26.0 ± 3.1
Women: 23.0 ± 2.4 Polar S410 + chest
HR strap Age, sex, height,
weight, and
physical activity
level
Resting HR: supine
position Indirect calorime-
try: ParvoMedics
TrueMax 2400
Treadmill:
individual ramp
running protocol
(individual start,
increase 1%
incline per min)
T test and Pear-
son’s r
Esco etal. 2011
[37]50 active men 24.0 ± 5.1 Polar F11 + chest
HR strap Age, sex, height,
weight, and
physical activity
level
Resting HR: supine
position Indirect calorime-
try: ParvoMedics
TrueOne 2400
Treadmill: Bruce
running protocol
(speed and incli-
nation increase
each 3min)
T test, Pearson’s r
and Bland–Altman
Esco etal. 2014
[40]20 female soccer
players 21.5 ± 1.7 Polar FT40 + chest
HR strap Age, sex, height,
weight, and
physical activity
level
Resting HR: 5min
supine position Indirect calorime-
try: ParvoMedics
TrueOne 2400
Treadmill: Bruce
running protocol
(speed and incli-
nation increase
each 3min)
Bland–Altman and
MAPE
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
1583
Validity of VO2max Estimated by Wearables
Table 1 (continued)
References Participants Age (years) Wearable device.
HR assessment Setup information VO2max estimation Reference standard VO2max protocol Statistical analysis
Freeberg etal.
2019 [46]30 healthy, men
(17) and women
(13)
21.7 ± 3.1 Fitbit Charge 2.
Wrist-measured
HR (PPG)
Not specified Exercise test:
2 × 10min at
highest intensity
possible
Indirect calorime-
try: ParvoMedics
TrueOne 2400
Treadmill: individ-
ual ramp running
protocol (4–7
mph, increase
1% incline per
min) + verifica-
tiontest
ANOVA, Pearson’s
r, MAPE, Bland–
Altman and ICC
Klepin etal. 2019
[25]65 healthy men
(27) and women
(33)
31.0 ± 7.3 Fitbit Charge 2.
Wrist-measured
HR (PPG)
Age, sex, handed-
ness, height, and
weight
Exercise test:
3 × 15min at
comfortable pace
Indirect calorim-
etry: Cosmed Treadmill: ramp
running protocol
(5 mph, increase
by 0.75 MET per
min)
Bland–Altman and
MAPE
Kraft and Dow
2017 [30]16 healthy, men
(10) and women
(6)
22.4 ± 5.2 Garmin Forerunner
920XT + chest HR
strap
Height and weight Exercise test:
10min self-paced
run
Indirect calorime-
try: ParvoMedics
TrueOne 2400
Treadmill: Bruce
running protocol
(speed and incli-
nation increase
each 3min)
T test
Kraft and Dow
2018 [31]18 healthy, men
(12) and women
(6)
21.3 ± 2.2 Polar
RS300X + chest
HR strap
Age, height,
weight, sex, and
activity level
Resting HR: 5min
supine position Indirect calorime-
try: ParvoMedics
TrueOne 2400
Treadmill: Bruce
running protocol
(speed and incli-
nation increase
each 3min)
T test and Pear-
son’s r
Lowe etal. 2010
[51]32 active women 20.3 ± 1.9 Polar F6 + chest HR
strap Age, sex, height,
and weight Resting HR: 5min
sitting position Indirect calorime-
try: ParvoMedics Treadmill: Bruce
running protocol
(speed and incli-
nation increase
each 3min)
T test
Passler etal. 2019
[39]24 healthy, men
(13) and women
(11)
23.4 ± 2.1 Polar V800. Wrist-
measured HR
(PPG)
Not specified Resting test: 10min
supine position
(pretest), 3min
supine position,
3min standing
position
Indirect calorim-
etry: Metalyzer
3B-R3, Cortex
Treadmill:
ramp protocol
(7km·h−1,
increase by
0.5km·h−1 per
min)
T test, MAPE,
Bland–Altman and
ICC
Garmin Forerunner
920 XT. Wrist-
measured HR
(PPG)
Not specified Exercise
test: > 10min
self-paced run
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
1584 P.Molina-Garcia et al.
(± 23.03 vs. ± 17.97ml·kg−1·min−1). It was not possible to
make a comparison in the resting conditions due to only two
studies using PPG.
3.3 The Current State ofKnowledge inValidation
Protocols Relevant toInform Best‑Practice
Recommendations
Similar to the previous statements of the INTERLIVE con-
sortium [19, 20], we present and discuss the information
found in these studies divided into the six key domains to
take into consideration when designing validation protocols
of consumer wearables estimating VO2max (Fig.5).
3.3.1 Target Population
The total sample size studied was 403 participants (218
men and 185 women), with a mean sample per article of
29 participants. For future validation studies, we recom-
mend performing a priori sample size calculation following
the approach by Lu etal. [43], which uses the Bland–Alt-
man limit of agreement analysis. The required sample size
to obtain a power of 80–90% is calculated considering the
expected mean absolute difference between the index meas-
ure and the reference standard, the expected SD of this dif-
ference, and the maximum allowed difference predefined
by the researchers. It is advised to conduct a pilot study
to obtain this information directly from the devices to be
validated. If this is not feasible, our meta-analysis reveals
that the expected mean absolute difference in the resting
conditions is 2.30ml·kg−1·min−1 and the expected SD is
7.20ml·kg−1·min−1, whereas the expected mean absolute
difference in the exercise test is 1.32ml·kg−1·min−1 and the
expected SD is 4.03ml·kg−1·min−1. Regarding the maximum
allowed difference, there is no agreement on this size with
respect to relevance for performance, health promotion, or
clinical practice. In the second paragraph of the “Discus-
sion” section, we argue the potential meaningfulness of
the estimation errors by wearables considering previous
meta-analyses on VO2max changes and mortality risk. How-
ever, it is important to know that this maximum allowed
difference must be greater than the expected mean differ-
ence ± 1.96 × the expected SD. Thus, considering our meta-
analysis results, these values should be at least 16.41 and
9.22ml·kg−1·min−1 in the resting conditions and exercise
test, respectively. Raising the sample size will not affect the
estimated size of the limit of agreement but will provide
greater precision (i.e., tighter confidence bands around the
limit of agreement).
Participants from the included studies were adults with a
pooled age of 24.6 ± 5.7years old. However, children, ado-
lescents and older adults also use these wearable devices in
real life, and, therefore, we recommend that future validation
Table 1 (continued)
References Participants Age (years) Wearable device.
HR assessment Setup information VO2max estimation Reference standard VO2max protocol Statistical analysis
Snyder etal. 2019
[28]44 healthy, men
(22) and women
(22)
Men: 24.7 ± 5.4
Women: 25.0 ± 4.3 Polar V800 + chest
HR strap Age, sex, height,
weight, and
physical activity
level
Resting HR: 5min
supine position Indirect calorime-
try: ParvoMedics
TrueOne 2400
Treadmill: Bruce
running protocol
(speed and incli-
nation increase
each 3min)
ANOVA, Bland–
Altman and
Pearson’s r
Garmin Forerunner
230 + chest HR
strap
Age, sex, height,
weight, and
HRmax
Exercise test:
10min self-
paced run
Wagner etal. 2020
[42]23 healthy men 23.1 ± 2.5 Garmin GF5 Exercise test:
10min and 30s
all out run
Indirect calorim-
etry: Metalyzer
3B, Cortex
Treadmill: ramp
running protocol
(10km·h−1,
incline 5%,
increase by 2.5%
per min)
Bland–Altman and
ICC
ANOVA analysis of variance, HR heart rate, HRmax maximum heart rate, ICC intraclass correlation coefficient, MAPE mean absolute percentage error, MET metabolic equivalent, PPG photop-
lethysmography, VO2max maximal oxygen consumption
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
1585
Validity of VO2max Estimated by Wearables
studies include different age populations to ensure that the
validity is representative of the general population. Regard-
ing sex differences, Crouter etal. [27] found a remarkably
larger error when estimating VO2max in women compared to
men, while Snyder etal. [28] showed opposite results, with a
greater error in men compared to women. We suggest future
studies to test whether the validity of existing methods/algo-
rithms systematically differs according to sex.
In the risk of bias assessment, we identified that the
majority of articles (10 of 14) adequately delimited the target
population they wanted to study and nearly all participants
contributed with data to be included in the validity analysis.
Fig. 2 Risk of bias assessment
divided by domains
Fig. 3 Pooled bias and SE for wearables VO2max using resting con-
ditions (A) and exercise tests (B) relative to the reference standard.
A negative bias represents an underestimation and a positive bias
an overestimation of the VO2max estimated from wearables in com-
parison to the reference standard. CI confidence interval, SE standard
error, VO2max maximal oxygen consumption. *Heart rate was meas-
ured with chest strap. In the remaining articles not flagged with an
asterisk, heart rate was measured using photoplethysmography tech-
nology on the wrist
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
1586 P.Molina-Garcia et al.
Participants from the included studies were all physically
active people categorized as “healthy” or “active,” rec-
reational runners [29, 44] or soccer players [40]. In order
to have a wider representation of the general population,
VO2max estimations from consumer wearables should be
tested in further clinical populations such as old adults, indi-
viduals with more sedentary behaviors, with overweight/
obesity, or highly trained athletes. We, therefore, recom-
mend expanding the population included beyond healthy
young people (e.g., from very untrained sedentary people
to highly trained athletes), as well as to clearly define and
report the inclusion/exclusion criteria used to define these
target populations.
3.4 Reference Standard
All studies included indirect calorimetry through gas anal-
ysis as a reference standard of VO2max, as was previously
recommended [45]. In brief, indirect calorimetry measures
VO2 and VCO2 concentrations and calculates the respiratory
exchange ratio (RER), allowing for the obtainment of VO2max
while exercising [45]. The gas analysis systems used were
reported in all studies, where Parvo Medics was the most
popular brand, used in ten studies [27–31, 37, 38, 40, 44,
46], followed by Cosmed [25, 47] and Metalyzer [39, 42],
with two studies each. Although the validity and reliability
of indirect calorimetry systems may seem obvious, available
devices are not always reliable [48, 49] and only one of the
included studies provided a reference with regards to the
validity within the study [29]. Similarly, only two studies
included in this review specified whether the gas exchange
was recorded breath by breath [39, 42]. Furthermore, none of
the included articles reported whether the gas analyzer used
both VO2 and VCO2 for VO2max assessment, even though
it is known that systems without CO2 sensors decrease the
precision and should be treated with caution [50]. Lastly,
four studies [39, 42, 44, 47] did not clarify whether the
device was calibrated [45], and we recommend that a proper
calibration process according to the manufacturer’s instruc-
tions be performed before the VO2max assessment. We urge
Fig. 4 Bland–Altman meta-analysis for the comparison of wearable-
derived VO2max using resting conditions and exercise tests with the
reference VO2max. The y-axis is the bias between the wearable and
reference VO2max (wearable − reference), with positive values indicat-
ing an overestimation and negative values an underestimation by the
wearable. The x-axis is the mean VO2max between the wearable and
reference. CI confidence interval, VO2max maximal oxygen consump-
tion. *Heart rate was measured with chest strap. In the remaining arti-
cles not flagged with an asterisk, heart rate was measured using pho-
toplethysmography technology on the wrist
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
1587
Validity of VO2max Estimated by Wearables
authors and developers to improve transparent reporting by
including at a minimum the brand used, the type of record-
ing technology (e.g., breath by breath or mixing chamber),
and previous validity/reliability of the instruments.
3.5 Index Measure
Within the included studies in this review, eight validated
the VO2max estimations of Polar® devices (models: A300,
S410, F11, FT40, F6, RS300X, and two V800) [27, 28, 31,
37, 39, 40, 47, 51], five validated Garmin® devices (models:
Fenix 3, Fenix 5X, Forerunner 920 XT, and GF5) [29, 30,
39, 42, 44], and two validated Fitbit® devices (models: two
Charge 2) [25, 46]. However, several other brands currently
claim to provide VO2max estimations, such as Apple, Tom-
Tom, Huawei, Suunto, Withings, and Coros (Supplementary
Material 8; see the electronic supplementary material). Con-
sidering that scientific validation of these devices is lacking,
we suggest future validity studies on these remaining brands
in order to improve transparency.
Three out of the 14 included studies did not follow an
ecological validity procedure [28, 29, 44], defined as a
validation process that resembles the use of the device in
the consumer’s real life. Two of the studies introduced bias
when including the setup information, an aspect that will be
discussed in the “Testing Protocols and Conditions” section
[28, 44], while one study did not place the device in an eco-
logical manner according to manufacture instructions [29].
Regarding the ecological placement, Anderson etal. [29]
fixed the device to the wrist with additional tape, and this is
not recommended since it may artificially improve the preci-
sion of the HR readings through PPG, biasing the validity
of the device in ecological settings. Overall, we recommend
that wearable devices be worn on ecological body locations
in accordance with the manufacturer’s instructions, and this
location should be adequately described within the methods.
If multiple wrist-worn devices are being tested, a maximum
of two devices per wrist should be used at the same time,
with placement being randomly counterbalanced between
participants.
Apart from the wrist-worn wearables, nine devices incor-
porated a chest strap to record HR during the VO2max esti-
mation [28, 30, 37, 38, 40, 44, 47]. Chest-strap technology
has been the most used method for HR monitoring in the
past. Moreover, it is widely accepted as a valid and reli-
able method to measure HR in free-living conditions, but
it presents limitations in 24h recording over multiple days.
Recently, many wearables are built with the possibility to
measure HR at the wrist using the PPG technology, which
allows longer recording time and a more comfortable meas-
urement by not incorporating additional devices along with
the wrist bracelet (e.g., chest strap). A recent meta-analysis
Fig. 5 Six domains and corresponding variables of interest identified as being of importance in the validation of consumer wearable estimation
of VO2max. VO2max maximal oxygen consumption
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
1588 P.Molina-Garcia et al.
has also revealed an acceptable validity of the PPG technol-
ogy during treadmill running and walking (mean difference
− 0.51bpm; 95% CI − 1.60 to 0.58bpm), yet an underesti-
mation when performing endurance sports (mean difference
− 7.26bpm; 95% CI − 10.46 to − 4.07bpm) [52]. There-
fore, the type of HR measurement is relevant and should
be reported in the validation protocols. Future research is
necessary to determine whether the VO2max estimation is
more accurate using the HR obtained by PPG or chest strap.
Furthermore, the validity of HR measures from wearables
should be tested before being used in the VO2max estimation
following the recently published recommendations by the
INTERLIVE consortium [19].
3.6 Testing Protocols andConditions
3.6.1 Reference Standard
All of the included studies tested VO2max in laboratory con-
ditions. The two previous expert statements of the INTER-
LIVE consortium on step count and HR provided recom-
mendations for semi-free-living and free-living conditions
besides the laboratory setting to test the ecological validity
[19, 20]. However, reference VO2max is still recommended
to be performed in laboratory conditions, and, therefore,
the free-living and semi-free-living conditions do not apply
in this context. Regarding the type of activity, all included
studies applied treadmill running protocols. It is known that
running protocols may provide small differences in VO2max
in comparison to cycle protocols [53], and, therefore, our
recommendation is to incorporate protocols that are as close
as possible to the type of activity for which the consumer
wearable has been designed.
In regards with the work rate progression, some protocols
gradually increased the speed [25, 39], the treadmill incli-
nation [27, 42, 46], or both intensity conditions within the
protocol [28–31, 40, 41, 44, 47, 51]. Five studies used ramp
protocols [25, 27, 39, 42, 46] in which work rate increases
more gradually (e.g., each 30–60s), while the remainder
studies included blocks of 2 [44] or 3min [28–31, 37, 40,
47, 51]. It seems that VO2max does not vary whether tread-
mill inclination or speed increase is used [53]. Likewise,
the use of a ramp versus a more accentuated increase in
the work rate does not affect the VO2max measure, although
each progression has pros and cons depending on the tar-
get population and whether treadmill or cycle ergometer is
used [54]. We recommend selecting an appropriate work rate
progression according to the type of population in which
the consumer wearable is intended to be validated and the
selected physical activity (e.g., running or cycling).
Maximal graded exercise testing requires participants to
terminate the test at volitional fatigue, and accepted crite-
ria exist to ensure that maximal VO2 during the test was
reached. For more information, we refer readers to chapter4
of the American College of Sports Medicine’s (ACSM’s)
Guidelines for Exercise Testing and Prescription, in which a
detailed description of test termination criteria can be found
[7]. Among the included studies, five did not consider at
least two maximum-effort criteria apart from voluntary
exhaustion and are likely to have measured VO2peak instead
of VO2max [25, 30, 31, 39, 44]. In the last years, an alterna-
tive/complementary solution named “verification phase”
has been proposed, which includes an extra effort lasting
between 2 and 3min at a supramaximal work rate (i.e., 110%
of maximum power) after the test termination to corroborate
the results [55]. This approach was only followed by Free-
berg etal. [46] and may be an interesting method to use in
future validation protocols.
A maximal graded exercise test normally requires several
standardized conditions to ensure that the participants reach
their true VO2max. Five out of the 14 included articles con-
sidered at least some of these standardized conditions before
the exercise testing [27, 29, 38–40], whereas the remainder
did not report this information. The INTERLIVE consor-
tium recommends taking into account the following stand-
ardized conditions when measuring the VO2max reference
standard: caloric uptake, caffeine or alcohol consumption,
intensive sports activities, medications, and an appropriate
warm-up (e.g., 5–10min of light-intensity aerobic exercise
and dynamic stretching) before commencing the exercise
test [7, 53].
3.6.2 Wearable Device
Included studies that estimated VO2max from a resting test
were Polar devices and the test used was the patented “Polar
fitness test” [56]. Polar devices record the resting HR and
heart rate variability (HRV) via Polar chest strap or the PPG
technology incorporated into the device and use these data to
estimate VO2max [57]. This protocol slightly differed based
on the wearable model, but always ranged from 5 to 10min
in a supine position (e.g., Polar A300, FT40, and F6), while
only one of the included models additionally added a few
minutes in a standing position (e.g., Polar V800). On the
other hand, only Garmin and Fitbit were the brands that used
exercise testing. The Fitbit exercise test consists of a run
at a comfortable pace for at least 10min while the GPS is
being recorded [58]. Garmin devices offer different meth-
ods to estimate VO2max depending on three types of activity:
running, cycling, or walking [59]. However, only the run-
ning protocol was used in all studies included in this review
[28–30, 42, 44], requiring a run of at least 10min, while
recording the GPS signal and HR data (through PPG tech-
nology or chest strap). Garmin’s instructions recommend an
intensity of at least 70% of the user’s maximal HR for the
entire exercise, which can be either estimated or manually
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
1589
Validity of VO2max Estimated by Wearables
input by the user [59]. Overall, we recommend researchers
systematically follow the manufacturer’s recommendations
when estimating VO2max from the wearable device among
study participants.
Some of the included wearable devices require a previ-
ous setup in which personal data such as age, sex, height,
weight, or physical activity level are recorded to improve the
accuracy of the VO2max estimation. Only two of the included
studies did not specify whether previous setup information
was input prior to commencing the validation protocol [39,
46], while the remainder of the studies recorded some basic
information. As a general recommendation, all the setup
information required by the device should be included
and reported, and this should be similar to the information
customers are provided outside of a research context. For
instance, both Snyder etal. [28] and Carrier etal. [44] intro-
duced the maximum heart rate (HRmax) obtained from the
reference standard test into the consumer wearables, which
is not ecological since few users have HRmax data from a
maximal graded exercise test in laboratory conditions.
3.7 Data Processing
3.7.1 Reference Standard
Indirect calorimetry for either mixing-chamber or breath-
by-breath technology requires several decisions on data pro-
cessing while conducting VO2max tests. A major factor for
removing variability in indirect calorimetry is the time and
breath averages used to estimate VO2max. Only three [25, 27,
46] of the studies included in this review reported this rel-
evant information. Following Robergs etal. [26] recommen-
dations, between 15 and 30s time averages and 15-breath
running averages should be used to have a reasonable reduc-
tion in data variability without losing relevant physiological
information. For researchers implementing digital filters, a
low cut-off frequency of 0.04Hz is recommended [26].
3.7.2 The Time Interval Between Evaluations
With regards to wearable devices, modifying data process-
ing is not possible since the wearables directly compute the
VO2max using algorithms that are usually proprietary infor-
mation and the exact equations are not disclosed. An impor-
tant consideration, however, is the time interval between
both assessments, since the fatigue after the maximal exer-
cise test may affect the wearable VO2max estimation. Since
the resting methodology is conducted in resting conditions,
these wearable protocols can be performed before the refer-
ence standard protocol without influencing either test. This
should not be performed in the opposite order, since the
maximal test required for the reference standard could affect
the resting HR or HRV. Concerning the wearable estimations
based on the exercise test, 24–48h between tests is rec-
ommended to ensure optimal recovery from high-intensity
exercise and avoid associated muscle fatigue hampering the
performance [60]. Furthermore, randomization or counter-
balancing the order of the wearable and laboratory tests is
important to control the potential carryover effects. Five of
the included studies in this review either did not meet this
time-interval criterion or did not report any information [25,
28, 29, 39, 42], and none mentioned any randomization or
counterbalancing strategy, which is an aspect to consider in
future validation studies.
3.8 Statistical Analysis
The Bland–Altman limits of agreement analysis is the
most popular method used in validation studies and has
been widely accepted as the most appropriate type of sta-
tistical analysis in these types of studies [61, 62]. In brief,
Bland–Altman analysis provides both the systematic error
(i.e., bias or average difference between methods) and the
random error or precision (i.e., 95% limit of agreement of
the systematic error), thus providing valuable information
for the comparison of the wearable devices to the reference
standard. The lower and upper bound of the limits of agree-
ment provides an estimate in which 95% of future obser-
vations of the differences in VO2max between the wearable
device and a criterion reference assessment are expected
to fall. In addition, the Bland–Altman plots represent the
individual difference between methods against the mean of
the methods, providing visual information on other relevant
dimensions of agreement, such as heteroscedasticity (a trend
to increase/decrease the error between methods as the mag-
nitude of the measurement increases). Additionally, percent-
age error measures, such as the mean absolute percentage
error (MAPE), represent a helpful option to report the error
of the device in an easy-to-understand manner [63]. There-
fore, we recommend reporting percentage error measures
complementary to the limit of agreement analysis. In the
risk of bias assessment, we detected that five studies did
not apply an appropriate analysis of agreement between the
wearable devices and the reference standard, since they only
performed mean difference (t test or analysis of variance
[ANOVA], but did not report the limits of agreement or the
Bland–Altman plots) or Pearson correlation analyses [27,
29–31, 47, 51]. Among the statistical tests used, Bland–Alt-
man [25, 28, 37, 39, 40, 42, 44, 46], t test [27, 29–31, 37–39,
44], and Pearson’s r [27–29, 31, 37, 44, 46, 47] were the
most popular tests, with eight studies using each of these
analyses, followed by MAPE in five studies [25, 39, 40,
44, 46] and intraclass correlation coefficient [39, 42, 46] or
ANOVA [28, 46, 47] in three studies each.
The last point to consider is the contextual validity of
wearable devices in estimating VO2max, which should be
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
1590 P.Molina-Garcia et al.
considered within the statistical analysis. For instance, if
a wearable device is designed to monitor VO2max changes
that improve users’ health, the systematic and random errors
should be critically analyzed to ensure that the device is
capable of detecting individual changes, which are con-
sidered clinically significant in the scientific literature. We
have already proposed in the “Methods” section that 3.5 and
1.75ml·kg−1·min−1 might be potential thresholds since both
are normal VO2max changes in the general population and
have been associated with health improvements. Therefore,
companies should report the level of error in a transparent
manner according to the purpose of the device and the target
population. This would guide researchers in the statistical
analysis and the interpretation of the results.
3.9 Recommended Validation Protocol
Based on the abovementioned state of knowledge and the
critical discussion between the members of the INTERLIVE
consortium, we present best-practice recommendations for
validation protocols of VO2max derived from consumer wear-
able devices in Table2. Furthermore, a checklist is provided
in Table3, including the items to be considered when plan-
ning validation protocols of VO2max consumer wearables. A
graphical overview of the six domains to consider in these
validation protocols is presented in Fig.5.
4 Discussions, Future Directions,
andStatement
In the present article, we combined a systematic review and
meta-analysis with an expert statement aiming (1) to pro-
vide a summary of the validity of VO2max estimations by
consumer wearables that use different methods/algorithms
and (2) to provide recommendations for future validation
studies. Our meta-analysis suggests that consumer weara-
bles using exercise tests provided a more accurate estima-
tion of VO2max in comparison to consumer wearables using
resting tests. Overall, the wearables using exercise tests
to estimate VO2max had a systematic error close to zero
(−0.09ml·kg−1·min−1) in comparison to maximal graded
exercise tests using indirect calorimetry in laboratory con-
ditions. However, the random error observed in both types
of methods was still large, i.e., limits of agreements span
of ± 15.24 (95% CI − 22.18 to 26.53) and ± 9.83 (95% CI
− 16.79 to 16.61) ml·kg−1·min−1 for the resting and exercise
tests, respectively. Consequently, even if this random error
was markedly smaller in the exercise-based estimations, it
is still a large error when estimating VO2max at an individual
level.
We are unaware of any well-established and accepted
estimation error to strongly indicate when the validity of a
wearable is acceptable or not. Our aim here was to inform
the public about the observed estimation errors based on
existing literature. It is ultimately up to the users to con-
sider whether the error is good enough for their specific
purposes. Just to put into context the potential meaning-
fulness of estimation errors observed in VO2max, we need
to consider that previous meta-analyses have reported that
increases in VO2max of 1.75–3.5ml·kg−1·min−1 are associ-
ated with a lower risk of all-cause mortality and incidence
of coronary heart disease or cardiovascular disease [5, 64].
Therefore, systematic and random errors in the estimation
by wearables beyond the range of 3.5ml·kg−1·min−1 will be
missing clinically relevant changes. Reliability is also an
important concept to understand the quality of the weara-
bles estimates; however, only three of the included studies
evaluated it [40, 41, 47]. Overall, good test–retest reliabil-
ity of wearable VO2max has been reported with r and intra-
class correlation coefficient (ICC) values above 0.90, but
further studies using a more recommendable approach (i.e.,
Bland–Altman limits of agreement) are needed to confirm
that wearable VO2max is reliable. Given the lack of evidence
regarding reliability, caution should be paid when wearables
are used for testing individual changes for either research,
clinical, or sports purposes. On the other hand, the estima-
tion errors of the exercise-based algorithms at the group
level show a high level of accuracy. This fact allows digital
phenotyping of cardiorespiratory fitness using wearables at
a population level, which opens new opportunities for fitness
monitoring at regional, national, or global levels. We cannot
determine the number of people for which the exercise-based
algorithms are accurate, but considering our results come
from 244 participants, we can establish this population cut-
off point for now.
In order to better understand the different errors observed
in the two types of estimation methods, it is important to
discuss how the different brands estimate VO2max through
different methodologies. Polar devices use resting HR,
HRV, gender, age, height, body weight, and self-reported
physical activity to estimate VO2max. The company explains
in a white paper that they used data from several valida-
tion studies to develop an artificial neural network that
calculates VO2max through the fitness test [65]. They claim
that the mean error of the prediction varies between 8%
(3.7ml·kg−1·min−1 approximately) and 15% compared with
laboratory test. Our results reveal an assumable systematic
error of 2.17ml·kg−1·min−1, but an overly wide random error
span of ± 30.48ml·kg−1·min−1. Polar claims the main benefit
of the Polar fitness test is that it is “easy, safe and convenient
for setting a baseline and tracking relative progress” [57].
We agree that a test in resting conditions is very convenient,
feasible, and safe and, therefore, a good solution when more
valid methods are not feasible. However, based on the wide
random error observed in the meta-analysis, we would not
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
1591
Validity of VO2max Estimated by Wearables
Table 2 The proposed best-practice protocols for the validation of wearable-derived VO2max
Domain Variable Protocol consideration Reporting consideration
Target population Population If purpose is to validate wearable-derived VO2max for the general healthy population, a
broad heterogeneous sample should be used
If purpose is to use wearables in specific clinical applications, validation should be
performed in homogenous samples
Report the inclusion/exclusion criteria defining the target
population and recruitment methodology and provide
basic demographic information (e.g., age, height, weight,
or BMI)
Age Validation protocols targeting a general healthy population should include the main age
ranges: children (< 12years), adolescents and adults (13–64years), and older adults Average and range of sample age should be reported
Sex Include an equal sample of males and females within the study The number of female and male participants should be
reported
Sample size For those studies aimed at testing the accuracy of a given device, a sample size calcula-
tion should be performed based on the previously published data according to Lu
etal.[43]. If no previous data are available or this is not the focus of the evaluation,
we advise to include a minimum of 15 participants per age group according to previ-
ously published recommendations on wearables-derived health measures [19, 20]
Describe the sample size calculation if included
If sample size calculation is not feasible, cite previous
literature supporting the inclusion of a recommended
sample size
Describe the flow of sample size recruited and analyzed
Reference standard Indirect calorimetry The gold standard for the assessment of VO2max is a maximal graded exercise test,
performed in laboratory conditions with indirect calorimetry [7]
Any brand of metabolic cart is accepted when reporting validity and reliability, as well
as measuring both VO2 and VCO2 during expiration
The metabolic cart should be properly calibrated before the VO2max assessment accord-
ing to manufacturer’s instructions
Indicate if indirect calorimetry was used
Report the metabolic cart used, the type of recording
technology (e.g., breath-by-breath), and whether the
metabolic cart used is valid and reliable
Describe the calibration process of the metabolic cart
Index measure Wearable devices Consumer wearables should be worn in ecological body locations in accordance with
the manufacturer’s instructions. If wrist worn, a maximum of 2 devices per wrist
should be used at the same time, with placement being randomly counterbalanced
between participants
Wearable devices can measure HR with PPG and/or chest-strap technology, and this
may have an impact on the VO2max estimation
Report the placement of the device and information on
order of placement if more than one wrist worn device is
used
Specify whether HR was recorded with PPG on wrist/arm
(or others) or chest-strap technology
Testing protocols
and conditions
for both refer-
ence and index
measure
Maximal graded
exercise testing
with indirect
calorimetry
The accepted protocol to assess VO2max is a maximal graded exercise testing evaluated
in laboratory conditions
Maximal test requires participants to perform to the point of volitional fatigue, and at
least two accepted criteria are recommended to ensure that participants are reaching
the maximum effort during the tests. The ACSM proposes several maximum-effort
criteria that can be used [7]
A verification phase after the maximal test is recommended to compare both VO2max
results. Schaun [55] provides an update of the literature on how to perform this veri-
fication phase
Any type of exercise testing is accepted (e.g., walking, running, or biking) as long as
it adapts to the type of activity in which the consumer wearable is intended to be
validated
In populations unable to perform maximal test, submaximal exercise-based equations
might be an alternative to predict VO2max, since overall these have demonstrated a
moderate to strong relationship with maximal tests. However, authors should select
the most appropriate equation for their target population [9, 70]
Report whether maximal or submaximal exercise test is
being used. In the case of submaximal test, provide a
rationale of its implementation and specify the exercise-
based equations used
In maximal exercise test, report the need for reaching voli-
tional fatigue and indicate the maximum-effort criteria
included (at least two criteria)
Report the type of exercise testing used as well as its
characteristics (e.g., increase in the ramp inclination in
treadmill tests or power increase in cycle-ergometer tests)
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
1592 P.Molina-Garcia et al.
Table 2 (continued)
Domain Variable Protocol consideration Reporting consideration
Standardized
conditions before
the reference and
index measure
Participants should not consume a significant caloric uptake at least 2h before the
exercise test
No caffeine, similar stimulants, or alcohol should be consumed 24h before the exercise
test
No intensive sports activities should be performed 48h before the exercise test
Participants should not take any medication that may alter the normal HR response to a
maximal exercise
The exercise test should begin with at least 2–3min warm-up
Report the standardized conditions followed by participants
Describe the warm-up characteristics
Wearable device
set up Follow the manufacturer’s instructions for the VO2max estimation protocol
Provide all the information required by the device, since in some cases this is used to
improve the VO2max estimation
If the device has the option to select a specific exercise mode (i.e., indoor running,
cycling, walking, etc.), choose the mode that best reflects the activity that is going to
be performed
In those wearable devices using GPS data, it is recommended to perform the test out-
door to ensure a proper GPS connection
Report the device model and version
Report what demographic details are input into the device
per participant for initiation
Report what mode (if any) is used during each activity (i.e.,
indoor running, cycling, walking, etc.)
If GPS is used, indicate that the satellite connection was
checked before the exercise test
Data processing Indirect calorimetry
processing If a time average is used to reduce variability in the indirect calorimetry data, typically
this should be between 15 and 30s [26]
If a breath average is used, a 15-breath running average is recommended [26]
Confirm that the maximum-effort criteria were met when interpreting the VO2max
values
Report the time-averaged or breath-averaged sampling used
Report whether maximal or peak VO2 is being assessed
Detail the data processing conducted in the VO2max inter-
pretation
Time interval
between evalua-
tions
If resting conditions are used for wearable VO2max estimation, no time interval is
needed before the reference VO2max test is performed
If the wearable test involves exercising, between 24 and 48h is recommended to ensure
an effective muscle recovery. If the maximal test is evaluated first, a time interval
between 48 and 72h is recommended [7]
Report the time interval between both assessments
Statistical analysis Statistical tests To assess device accuracy, the following statistical tests should be performed:
1. Bland–Altman with limits of agreement
2. Least product regression of the difference against the means
3. MAPE
Subgroup analysis is encouraged if sample size allows. (e.g., sex, age category, ethnic-
ity, BMI)
Include Bland–Altman plots for a visual inspection of the
validity results
Binary conclusions about the validity of the device should
not be made if a formal sample size analysis has not been
conducted
ACSM American College of Sports Medicine, BMI body mass index, HR heart rate, MAPE mean absolute percentage error, PPG photoplethysmography, VO2max maximal oxygen consumption
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
1593
Validity of VO2max Estimated by Wearables
advise users to rely on the estimated VO2max from resting
conditions, and future efforts to improve this methodology
are required.
Fitbit and Garmin use the algorithms developed by First-
beat Technologies in the VO2max estimation [29, 44, 46].
This method uses the following calculation steps [66]: (1)
logging of personal information (at least age), (2) an exer-
cise test with the wearable measuring HR and speed, (3)
HR data are segmented to different zones and the reliability
of these segments is calculated, and (4) the most reliable
data segments are used to estimate VO2max by using linear
or nonlinear dependency between HR and speed data. The
white paper published by Firstbeat stated that this estima-
tion had 5% MAPE for running, 8% for cycling, and 6% for
walking against indirect calorimetry VO2max in laboratory
settings [66]. Four studies in this systematic review reported
MAPE analyses of Fitbit and Garmin devices in running
tests [25, 39, 44, 46], and results were always greater than
the 5% reported by Firstbeat, with values ranging from 8 to
10.2%. There are no standard thresholds to determine an
optimal MAPE, but previous validity studies of consumer-
based wearables considered ≥ 10% as an indicator of inac-
curacy, which are values close to those found in the exercise
protocols [67]. Although the systematic error we found in
the meta-analysis for these wearables using exercise tests is
negligible (i.e., 0.09ml·kg−1·min−1), the random error span
of ± 9.83ml·kg−1·min−1 represents a considerable range that
may consider its use inappropriate to adequately assess and
monitor VO2max changes. Nevertheless, this estimation meth-
odology is clearly superior to the resting approach with 2.08
and 10.82ml·kg−1·min−1 less systematic and random error,
respectively. By removing articles prior to 2017, the resting
condition demonstrated an improvement in the accuracy of
0.51ml·kg−1·min−1. This analysis supports the notion that
new devices and/or algorithms are providing more accurate
estimates. Nevertheless, results from this article should
encourage developers to opt for exercise methodologies for
a more accurate VO2max estimation.
This article has detected several weaknesses in the valida-
tion process, which highlights the need for further and more
rigorous studies. Future validation studies should consider
the best-practice recommendations provided in this article
by the INTERLIVE consortium in the six main domains.
Our review has detected that the validity of wearables has
been tested only in healthy and physically active people with
a narrow age range (i.e., 25 ± 6years). A recent systematic
review identified several determinants of cardiorespiratory
fitness such as sex, age, education, socioeconomic status,
ethnicity, body mass index (BMI), body weight, waist cir-
cumference, body fat, resting HR, C-reactive protein, smok-
ing, alcohol consumption, and physical activity level [68].
Future validity studies should include participants across the
spectrum of some of these influencing factors to determine
how the wearable VO2max performs in different populations.
Moreover, the reference standard and its associated protocol
and data processing were, without a doubt, the most critical
point in terms of risk of bias in the included studies. There-
fore, future studies should improve the indirect calorimetry
protocols used according to the current exercise testing
guidelines.
Regarding the wearable devices, greater transparency
from companies regarding not only the algorithms but also
the data used to estimate VO2max would be desirable (yet
limited by proprietary issues). This would help research-
ers to better control variables during validation protocols.
For instance, if running speed and inclination are used in
the estimation, then the quality of GPS signal, track maps,
and altimeter sensors should be key components to consider
in validation studies. HR seems to provide key data in the
VO2max estimation, and a great proportion of the consumer
wearables in this review included chest strap for the HR
measurement instead of PPG. Overall, our results in the
meta-analyses demonstrated a greater bias and limit of
agreement in those devices using PPG compared to chest
strap. This is a somewhat expected finding since the meas-
urement error of the chest strap seems minimal compared
to electrocardiogram monitoring [69]. However, since
wearing chest straps is uncomfortable for many people
and the greater acceptability in the general population of
HR monitoring via PPG (usually placed on the wrist, i.e.,
smartwatches and bracelets), it is important that future valid-
ity studies use PPG technology and aim to obtain accurate
VO2max estimations with it. In a previous INTERLIVE arti-
cle, we discussed several factors affecting the accuracy of
PPG technology, such as skin tone, motion artifacts, contact
pressure, and ambient temperature [19]. Recommendations
from this article should be considered to ensure best prac-
tice in the validity, testing, and reporting of PPG-based HR
wearables estimating VO2max. Lastly, all available literature
estimated VO2max while running. Thus, future validity stud-
ies are needed in other activities, such as cycling or walking,
to cover a broader range of activities.
The statistical analysis used in the available validity
studies was often inappropriate, and consequently, future
protocols should use the statistical approaches considered
appropriate in validation studies. We recommend using the
Bland–Altman limits of agreement as the main analysis and
some percentage error (e.g., MAPE) as complementary and
informative information. Overall, the application of the
best-practice recommendations from the INTERLIVE con-
sortium would be beneficial for stakeholders by ensuring a
more valid and transparent metric derived from their devices
as well as for users who would receive more accurate and
reliable information about their VO2max level and, therefore,
their health status.
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
1594 P.Molina-Garcia et al.
5 Conclusion
This systematic review and meta-analysis from the INTER-
LIVE consortium summarizes the validity of VO2max esti-
mated from consumer wearables and provides best-practice
recommendations for future validation protocols. The meta-
analysis suggests that the estimation of VO2max by wearables
that use exercise-based algorithms provides higher accuracy
than those based on resting methods. The exercise-based
estimation seems to be optimal for application at the popula-
tion level, yet the estimation error at the individual level and,
therefore, use for sport/clinical purposes still needs further
improvement. The INTERLIVE network hereby provides
best-practice recommendations to be used in future protocols
Table 3 The INTERLIVE
checklist to be considered
for the validation protocol of
wearable to estimate maximal
oxygen consumption (VO2max)
See the Table2 for more detailed information about each item
INTERLIVE Towards Intelligent Health and Well-Being Network of Physical Activity Assessment, MAPE
mean absolute percentage error, PPG photoplethysmography
Target population assessment
Age
Children (< 12years)
Adolescents (12–18years)
Adults (18–65years)
Older adults (> 65years)
Sex (equal sample of males and females)
Sample size
Calculated based on previously published or pilot study data
OR
If previous data is not available, sample of convenience (n ≥ 45 participants)
Reference standard
The gold standard is a maximal exercise test in laboratory conditions with indirect calorimetry
Any brand of metabolic cart is accepted and should be calibrated following manufacturer’s instructions
Index device assessment
Consumer wearables placed according to manufacturer’s instructions to be tested in ecological locations
Hear rate can be measured with both chest strap or PPG, and it should be reported which of them was
used
Testing protocols and conditions
Reference standard
To consider at least 2 maximal-effort criteria during the incremental test
A verification phase after the maximal test is recommended to corroborate the VO2max
Any type of exercise testing is accepted (e.g., walking, running, or biking) as long as it adapts to the type
of activity in which the consumer wearable is intended to be validated
Control the standardized conditions before the maximal exercise test
Consumer wearable
Follow the manufacturer’s instructions for the VO2max estimation protocol
Provide all the setup information required by the devices
If exercise mode is available, choose the one that best reflects the activity to be performed
Ensure an optimal GPS connection when this data is used
Processing
Reference standard
If VO2max is averaged within a time window, it is recommended to use a 15- to 30-s window
If a breath-by-breath average is used, a 15-breath running average is recommended
Confirm that the maximum-effort criteria were met when interpreting the VO2max values
Time interval between evaluations
In those wearables using resting conditions, no time interval is needed
In exercise conditions, an interval between 24 and 48h is recommended
Statistical analysis
Bland–Altman with limits of agreement
Least products regression of the differencesagainst the means
MAPE
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
1595
Validity of VO2max Estimated by Wearables
to move towards a more accurate, transparent, and compara-
ble validation of VO2max derived from wearables.
Supplementary Information The online version contains supplemen-
tary material available at https:// doi. org/ 10. 1007/ s40279- 021- 01639-y.
Declarations
Funding This research was partly funded by Huawei Technologies
Oy (Finland) Co. Ltd. A limited liability company headquartered in
Helsinki, Finland.
Conflict of interest None of the authors has any conflict of interest to
declare.
Data availability statement This systematic review has no original data
to provide. Most of the data have been reported within the main text
or supplementary material. The database used for the meta-analysis
and the R script for the Bland–Altman limits of agreement analysis is
available upon request to the corresponding authors.
Author contributions PM-G, HLN, and MS performed the systematic
review, screening, and data extraction. PM-G and AG analyzed the risk
of bias of included studies. PM-G, AG, and JCB performed the meta-
analysis. PM-G and FBO wrote the first draft of the manuscript. RA,
MH-R, JS, WB, SC, UE, LBS, BC, JCB, and AG critically reviewed
the manuscript. All authors read and approved the final manuscript.
Open Access This article is licensed under a Creative Commons Attri-
bution 4.0 International License, which permits use, sharing, adapta-
tion, distribution and reproduction in any medium or format, as long
as you give appropriate credit to the original author(s) and the source,
provide a link to the Creative Commons licence, and indicate if changes
were made. The images or other third party material in this article are
included in the article's Creative Commons licence, unless indicated
otherwise in a credit line to the material. If material is not included in
the article's Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will
need to obtain permission directly from the copyright holder. To view a
copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.
References
1. Tankovska H. Fitness & activity tracker—statistics & facts [Inter-
net]. Statistica. 2020 [cited 2021 Apr 16]. https:// www. stati sta.
com/ topics/ 4393/ fitne ss- and-a.
2. Strain T, Wijndaele K, Dempsey PC, Sharp SJ, Pearce M, Jeon
J, etal. Wearable-device-measured physical activity and future
health risk. Nat Med [Internet]. 2020;26:1385–91. http:// www.
nature. com/ artic les/ s41591- 020- 1012-3.
3. Brickwood KJ, Watson G, O’brien J, Williams AD. Consumer-
based wearable activity trackers increase physical activity par-
ticipation: systematic review and meta-analysis. JMIR mHealth
uHealth. 2019.
4. Althoff T, Sosič R, Hicks JL, King AC, Delp SL, Leskovec J.
Large-scale physical activity data reveal worldwide activity ine-
quality. Nature. 2017.
5. Ross R, Blair SN, Arena R, Church TS, Després JP, Franklin BA,
etal. Importance of assessing cardiorespiratory fitness in clini-
cal practice: a case for fitness as a clinical vital sign: a Scientific
Statement from the American Heart Association. Circulation.
2016.
6. Bassett DR, Howley ET. Limiting factors for maximum oxygen
uptake and determinants of endurance performance. Med Sci
Sports Exerc. 2000.
7. ACSM. ACSM guidelines for exercise testing and preescripción.
Am. Coll. Sport. Med. 2018.
8. Bennett H, Parfitt G, Davison K, Eston R. Validity of submaximal
step tests to estimate maximal oxygen uptake in healthy adults.
Sport Med. 2016;46:737–50.
9. Smith AE, Evans H, Parfitt G, Eston R, Ferrar K. Submaximal
exercise-based equations to predict maximal oxygen uptake
in older adults: a systematic review. Arch Phys Med Rehabil.
2016;97:1003–12. https:// doi. org/ 10. 1016/j. apmr. 2015. 09. 023.
10. Behind our Science | Polar Global [Internet]. [cited 2021 Apr 22].
https:// www. polar. com/ en/ scien ce.
11.• Garmin R&D expenses 2014–2020 | Statista [Internet]. [cited
2021 Apr 22]. https:// www. stati sta. com/ stati stics/ 10362 22/
garmin- randd- expen diture/.
12. Evenson KR, Goto MM, Furberg RD. Systematic review of the
validity and reliability of consumer-wearable activity track-
ers. Int J Behav Nutr Phys Act. 2015. https:// doi. org/ 10. 1186/
s12966- 015- 0314-1.
13. Straiton N, Alharbi M, Bauman A, Neubeck L, Gullick J, Bhindi
R, etal. The validity and reliability of consumer-grade activity
trackers in older, community-dwelling adults: a systematic review.
Maturitas. 2018.
14. Fuller D, Colwell E, Low J, Orychock K, Tobin MA, Simango
B, etal. Reliability and Validity of commercially available
wearable devices for measuring steps, energy expenditure, and
heart rate: systematic review. JMIR mHealth uHealth [Internet].
2020;8:e18694. http:// mheal th. jmir. org/ 2020/9/ e18694/.
15. Zhang Y, Weaver RG, Armstrong B, Burkart S, Zhang S, Beets
MW. Validity of Wrist-Worn photoplethysmography devices to
measure heart rate: a systematic review and meta-analysis. J.
Sports Sci. 2020.
16. O’Driscoll R, Turicchi J, Beaulieu K, Scott S, Matu J, Deighton K,
etal. How well do activity monitors estimate energy expenditure?
A systematic review and meta-analysis of the validity of current
technologies. Br J Sports Med. 2020;54:332–40.
17. Keadle SK, Lyden KA, Strath SJ, Staudenmayer JW, Freedson PS.
A Framework to evaluate devices that assess physical behavior.
Exerc Sport Sci Rev. 2019;47:206–14.
18. Welk GJ, Bai Y, Lee JM, Godino JOB, Saint-Maurice PF, Carr L.
Standardizing analytic methods and reporting in activity monitor
validation studies. Med Sci Sports Exerc. 2019;51:1767–80.
19. Mühlen JM, Stang J, Lykke Skovgaard E, Judice PB, Molina-
Garcia P, Johnston W, etal. Recommendations for determining the
validity of consumer wearable heart rate devices: expert statement
and checklist of the INTERLIVE Network. Br J Sports Med. 2021.
20. Johnston W, Judice PB, Molina García P, Mühlen JM, Lykke Sko-
vgaard E, Stang J, etal. Recommendations for determining the
validity of consumer wearable and smartphone step count: expert
statement and checklist of the INTERLIVE network. Br J Sports
Med. 2020.
21. Standards—Tagged “Health and Fitness”—Consumer Technology
Association® [Internet]. [cited 2021 Apr 23]. https:// shop. cta. tech/
colle ctions/ stand ards/ health- and- fitne ss.
22. Mokkink LB, Boers M, van der Vleuten CPM, Bouter LM, Alonso
J, Patrick DL, etal. COSMIN Risk of Bias tool to assess the
quality of studies on reliability or measurement error of outcome
measurement instruments: a Delphi study. BMC Med Res Meth-
odol. 2020;20:1–13.
23. Sterne JAC, Savović J, Page MJ, Elbers RG, Blencowe NS,
Boutron I, etal. RoB 2: a revised tool for assessing risk of bias in
randomised trials. BMJ. 2019.
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
1596 P.Molina-Garcia et al.
24. Higgins JPT, Thompson SG, Spiegelhalter DJ. A re-evaluation of
random-effects meta-analysis. J R Stat Soc Ser A Stat Soc. 2009.
25. Klepin K, Wing D, Higgins M, Nichols J, Godino JG. Validity of
cardiorespiratory fitness measured with fitbit compared to VO2max.
Med Sci Sports Exerc. 2019;51:2251–6.
26. Robergs RA, Dwyer D, Astorino T. Recommendations for
improved data processing from expired gas analysis indirect calo-
rimetry. Sport Med. 2010.
27. Crouter SE, Albright C, Bassett DRJ. Accuracy of polar S410
heart rate monitor to estimate energy cost of exercise. Med Sci
Sports Exerc United States. 2004;36:1433–9.
28. Snyder NC, Willoughby CA, Smith BK. Comparison of the
Polar V800 and the Garmin Forerunner 230 to predict
V
O2max. J
Strength Cond Res [Internet]. 2019;Publish Ah:1–7. https:// journ
als. lww. com/ 00124 278- 90000 0000- 95017.
29. Anderson JC, Chisenall T, Tolbert B, Ruffner J, Whitehead PN,
Conners RT. Validating the commercially available Garmin Fenix
5x wrist-worn optical sensor for aerobic capacity. Int J Innov Educ
Res. 2019;7:147–58.
30. Kraft GL, Roberts RA. Validation of the Garmin Forerunner
920XT Fitness Watch VO2peak Test. Int J Innov Educ Res [Inter-
net]. 2017;5:63–9. https:// ijier. net/ ijier/ artic le/ view/ 619.
31. Kraft GL, Dow M. Validation of the polar fitness test. Int J Innov
Educ Res [Internet]. 2018;6:27–34. https:// ijier. net/ ijier/ artic le/
view/ 893.
32. 16.1.3.2 Imputing standard deviations for changes from baseline
[Internet]. [cited 2021 Apr 24]. https:// handb ook-5- 1. cochr ane.
org/ chapt er_ 16/ 16_1_ 3_2_ imput ing_ stand ard_ devia tions_ for_
chang es_ from_ basel ine. htm.
33. Drevon D, Fursa SR, Malcolm AL. Intercoder reliability and
validity of WebPlotDigitizer in extracting graphed data. Behav
Modif. 2017.
34. Tipton E, Shuster J. A framework for the meta-analysis of Bland-
Altman studies based on a limits of agreement approach. Stat
Med. 2017;36:3621–35.
35. DerSimonian R, Laird N. Meta-analysis in clinical trials revisited.
Contemp Clin Trials. 2015.
36. Sterne JAC, Egger M, Smith GD. Systematic reviews in health
care: investigating and dealing with publication and other biases
in meta-analysis. Br. Med. J. 2001.
37. Esco MR, Mugu EM, Williford HN, McHugh AN, Bloomquist
BE. Cross-validation of the polar fitness testTM via the polar F11
heart rate monitor in predicting VO2max. J Exerc Physiol Online.
2011;14:31–7.
38. Lowe AL, Lloyd LK, Miller BK, McCurdy KW, Pope ML. Accu-
racy of polar F6 in estimating the energy cost of aerobic dance
bench stepping in college-age females. J Sports Med Phys Fit.
2010;50:385–94.
39. Passler S, Bohrer J, Blöchinger L, Senner V. Validity of wrist-
worn activity trackers for estimating VO2max and energy expendi-
ture. Int J Environ Res Public Health. 2019;16.
40. Esco MR, Snarr RL, Williford HN. Monitoring changes in VO2max
via the Polar FT40 in female collegiate soccer players. J Sports
Sci. 2014;32:1084–90. https:// doi. org/ 10. 1080/ 02640 414. 2013.
879672.
41. Esco MR, Mugu EM, Williford HN, McHugh AN, Bloomquist
BE. Cross-validation of the polar fitness testTM via the polar F11
heart rate monitor in predicting VO2max. J Exerc Physiol Online
[Internet]. 2011;14:31–7. https:// www. scopus. com/ in ward/ record.
uri? eid=2- s2.0- 84856 91577 1& partn erID= 40& md5= 310f9 53a3e
86839 1daeb 3a0fa 3faac e1.
42. Wagner M, Engel F, Klier K, Klughardt S, Wallner F, Wieczorek
A. Zur Reliabilität von Wearable Devices am Beispiel einer Pre-
mium Multisport-Smartwatch. Ger J Exerc Sport Res [Internet].
2020. https:// doi. org/ 10. 1007/ s12662- 020- 00682-7.
43. Lu MJ, Zhong WH, Liu YX, Miao HZ, Li YC, Ji MH. Sample size
for assessing agreement between two methods of measurement by
Bland–Altman method. Int J Biostat. 2016.
44. Carrier B, Creer A, Williams LR, Holmes TM, Jolley BD, Dahl S,
etal. Validation of Garmin Fenix 3 HR fitness tracker biomechan-
ics and metabolics (VO2max). J Meas Phys Behav. 2020;3:331–7.
45. Schoffelen PFM, Plasqui G. Classical experiments in whole-body
metabolism: open-circuit respirometry—diluted flow chamber,
hood, or facemask systems. Eur J Appl Physiol. 2018.
46. Freeberg KA, Baughman BR, Vickey T, Sullivan JA, Sawyer BJ.
Assessing the ability of the Fitbit Charge 2 to accurately predict
VO2max. mHealth [Internet]. 2019;5:39–39. http:// mheal th. amegr
oups. com/ artic le/ view/ 29481/ html.
47. Cooper KD, Shafer AB. Validity and reliability of the Polar
A300’s fitness test feature to predict VO2max. Int J Exerc Sci.
2019;12:393–401.
48. Cooper JA, Watras AC, O’Brien MJ, Luke A, Dobratz JR,
Earthman CP, etal. Assessing validity and reliability of rest-
ing metabolic rate in six gas analysis systems. J Am Diet Assoc.
2009;109:128–32.
49. Carter J, Jeukendrup AE. Validity and reliability of three com-
mercially available breath-by-breath respiratory systems. Eur J
Appl Physiol. 2002.
50. Macfarlane DJ. Open-circuit respirometry: a historical
review of portable gas analysis systems. Eur J Appl Physiol.
2017;117:2369–86. https:// doi. org/ 10. 1007/ s00421- 017- 3716-8.
51. Lowe AL, Lloyd LK, Miller BK, McCurdy KW, Pope ML. Accu-
racy of polar F6 in estimating the energy cost of aerobic dance
bench stepping in college-age females. J Sports Med Phys Fit
[Internet]. 2010;50:385–94. http:// www. ncbi. nlm. nih. gov/ pub-
med/ 21178 923.
52. Zhang Y, Weaver RG, Armstrong B, Burkart S, Zhang S, Beets
MW. Validity of Wrist-Worn photoplethysmography devices
to measure heart rate: a systematic review and meta-analysis. J
Sports Sci. 2020;38:2021–34. https:// doi. org/ 10. 1080/ 02640 414.
2020. 17673 48.
53. Beltz NM, Gibson AL, Janot JM, Kravitz L, Mermier CM, Dalleck
LC. Graded exercise testing protocols for the determination of
VO2max: historical perspectives, progress, and future considera-
tions. J Sports Med. 2016;2016:1–12.
54. Mezzani A. Cardiopulmonary exercise testing: Basics of method-
ology and measurements. Ann Am Thorac Soc. 2017.
55. Schaun GZ. The maximal oxygen uptake verification phase: a light
at the end of the tunnel? Sport Med Open. 2017;3.
56. Polar Fitness Test | Polar Blog [Internet]. [cited 2021 Apr 21].
https:// www. polar. com/ blog/ lets- talk- polar- polar- fitne ss- test/.
57. Polar Orthostatic Test. 2019 [cited 2021 Apr 16]. www. polar . com.
58. What is my cardio fitness score? [Internet]. [cited 2021 Apr 21].
https:// help. fitbit. com/ artic les/ en_ US/ Help_ artic le/ 2096. htm.
59. What is VO2max. Estimate and how does it work? | Garmin Sup-
port [Internet]. [cited 2021 Apr 21]. https:// suppo rt. garmin. com/
en- US/? faq= lWqSV lq3w7 6z5Wo ihLy5 f8.
60. Bishop-Fitzpatrick L, Mazefsky CA, Eack SM. The combined
impact of social support and perceived stress on quality of life
in adults with autism spectrum disorder and without intellectual
disability. Autism [Internet]. University of Wisconsin, Madison,
United States: SAGE Publications Ltd; 2018;22:703–11. https://
www. scopus. com/ inward/ record. uri? eid=2- s2.0- 85041 62515 4&
doi= 10. 1177% 2F136 23613 17703 090& partn erID= 40& md5=
8380e b3d3e 32bf5 f51dc 1ac5c bd7a6 af.
61. Martin Bland J, Altman DG. Statistical methods for assessing
agreement between two methods of clinical measurement. Lancet.
1986.
62. Zaki R, Bulgiba A, Ismail R, Ismail NA. Statistical methods
used to test for agreement of medical instruments measuring
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
1597
Validity of VO2max Estimated by Wearables
continuous variables in method comparison studies: a systematic
review. PLoS One. 2012.
63. Tayman J, Swanson DA. On the validity of MAPE as a measure
of population forecast accuracy. Popul Res Policy Rev. 1999.
64. Kodama S, Saito K, Tanaka S, Maki M, Yachi Y, Asumi M, etal.
Cardiorespiratory fitness as a quantitative predictor of all-cause
mortality and cardiovascular events in healthy men and women:
a meta-analysis. JAMA. 2009.
65. Polar Research and Technology (White Paper). Polar Fitness Test
[Internet]. 2019. https:// www. polar. com/ en/ scien ce/ white papers/
fitne ss- test.
66. Firstbeat. Automated fitness level (VO2max) estimation with heart
rate and speed data. © 2014 Firstbeat Technol. 2017;1–9.
67. Nelson MB, Kaminsky LA, Dickin DC, Montoye AHK. Validity
of consumer-based physical activity monitors for specific activity
types. Med Sci Sports Exerc. 2016.
68. Zeiher J, Ombrellaro KJ, Perumal N, Keil T, Mensink GBM, Fin-
ger JD. Correlates and determinants of cardiorespiratory fitness
in adults: a systematic review. Sport Med Open. 2019.
69. Gillinov S, Etiwy M, Wang R, Blackburn G, Phelan D, Gillinov
AM, etal. Variable accuracy of wearable heart rate monitors dur-
ing aerobic exercise. Med Sci Sports Exerc. 2017.
70. Ferrar K, Evans H, Smith A, Parfitt G, Eston R. A systematic
review and meta-analysis of submaximal exercise-based equations
to predict maximal oxygen uptake in young people. Pediatr Exerc
Sci. 2014;26:342–57.
Authors and Aliations
PabloMolina‑Garcia1,2 · HannahL.Notbohm3· MoritzSchumann3,4· RobArgent5,6,7·
MeganHetherington‑Rauth8· JulieStang9· WilhelmBloch3· SulinCheng3,4· UlfEkelund9· LuisB.Sardinha8·
BrianCauleld5,6· JanChristianBrønd10· AndersGrøntved10· FranciscoB.Ortega1,11,12
1 PROFITH (PROmoting FITness andHealth Through
Physical Activity) Research Group, Department ofPhysical
Education andSports, Faculty ofSport Sciences, University
ofGranada, Carretera de Alfacar s/n, 18071Granada, Spain
2 Physical Medicine andRehabilitation Service, Biohealth
Research Institute, Virgen de Las Nieves University Hospital,
Jaén Street, s/n, 18013Granada, Spain
3 Institute ofCardiovascular Research andSports Medicine,
Department ofMolecular andCellular Sports Medicine,
German Sport University, Cologne, Germany
4 Department ofPhysical Education, Exercise Translational
Medicine Centre, The Key Laboratory ofSystems
Biomedicine, Ministry ofEducation, andExercise, Health
andTechnology Centre, Shanghai Jiao Tong University,
Shanghai, China
5 Insight Centre forData Analytics, University College Dublin,
Dublin, Ireland
6 School ofPublic Health, Physiotherapy andSport Science,
University College Dublin, Dublin, Ireland
7 School ofPharmacy andBiomolecular Sciences, Royal
College ofSurgeons inIreland, Dublin, Ireland
8 Exercise andHealth Laboratory, CIPER, Faculdade de
Motricidade Humana, Universida de de Lisboa, Lisbon,
Portugal
9 Department ofSport Medicine, Norwegian School ofSport
Sciences, Oslo, Norway
10 Department ofSports Science andClinical Biomechanics,
Research Unit forExercise Epidemiology, Centre ofResearch
inChildhood Health, University ofSouthern Denmark,
OdenseM, Denmark
11 Faculty ofSport andHealth Sciences, University
ofJyväskylä, Jyvaskyla, Finland
12 Department ofBioscience andNutrition, Karolinska
Institutet, Huddinge, Sweden
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
1.
2.
3.
4.
5.
6.
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:
use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
use bots or other automated methods to access the content or redirect messages
override any security feature or exclusionary protocol; or
share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at
onlineservice@springernature.com
Content uploaded by Moritz Schumann
Author content
All content in this area was uploaded by Moritz Schumann on Jan 24, 2022
Content may be subject to copyright.
