ORIGINAL RESEARCH

published: 17 December 2021

doi: 10.3389/fphys.2021.782167

Novel Computerized Method for

Automated Determination of
Ventilatory Threshold and
Respiratory Compensation Point
Kyoung Jae Kim 1*, Eric Rivas 1, Brian Prejean 1, Dillon Frisco 2, Millennia Young 3 and
Meghan Downs 3
1
KBR, Houston, TX, United States, 2 JES Technologies, Houston, TX, United States, 3 NASA Johnson Space Center,
Houston, TX, United States

Introduction: The ventilatory threshold (named as VT1) and the respiratory compensation
point (named as VT2) describe prominent changes of metabolic demand and exercise
intensity domains during an incremental exercise test.
Edited by:
Gary W. Mack, Methods: A novel computerized method based on the optimization method was
Brigham Young University,
developed for automatically determining VT1 and VT2 from expired air during a progressive
United States
maximal exercise test. A total of 109 peak cycle tests were performed by members of
Reviewed by:
Stephen Seiler, the US astronaut corps (74 males and 35 females). We compared the automatically
University of Agder, Norway determined VT1 and VT2 values against the visual subjective and independent analyses
Glen E. Foster,
University of British Columbia of three trained evaluators. We also characterized VT1 and VT2 and the respective absolute
Okanagan, Canada and relative work rates and distinguished differences between sexes.
*Correspondence:
Kyoung Jae Kim Results: The automated compared to the visual subjective values were analyzed for
[email protected] differences with t test, for agreement with Bland–Altman plots, and for equivalence with
a two one-sided test approach. The results showed that the automated and visual
Specialty section:
This article was submitted to subjective methods were statistically equivalent, and the proposed approach reliably
Exercise Physiology, determined VT1 and VT2 values. Females had lower absolute O2 uptake, work rate, and
a section of the journal
ventilation, and relative O2 uptake at VT1 and VT2 compared to men (p ≤ 0.04). VT1 and
Frontiers in Physiology
VT2 occurred at a greater relative percentage of their peak VO2 for females (67 and 88%)
Received: 23 September 2021
Accepted: 18 November 2021 compared to males (55 and 74%; main effect for sex: p < 0.001). Overall, VT1 occurred
Published: 17 December 2021 at 58% of peak VO2, and VT2 occurred at 79% of peak VO2 (p < 0.0001).
Citation:
Kim KJ, Rivas E, Prejean B, Frisco D, Conclusion: Improvements in determining of VT1 and VT2 by automated analysis are
Young M and Downs M (2021) Novel time efficient, valid, and comparable to subjective visual analysis and may provide valuable
Computerized Method for Automated
Determination of Ventilatory information in research and clinical practice as well as identifying exercise intensity domains
Threshold and Respiratory of crewmembers in space.
Compensation Point.
Front. Physiol. 12:782167. Keywords: incremental exercise, noninvasive measurement, ventilatory threshold, respiratory compensation
doi: 10.3389/fphys.2021.782167 point, automated determination

Frontiers in Physiology | www.frontiersin.org 1 December 2021 | Volume 12 | Article 782167

Kim et al. Automated VT and RCP Detection

INTRODUCTION Exercise countermeasures are the only known way for

maintaining muscle mass, strength, and cardiorespiratory fitness
Examining the ventilatory profile during incremental exercise in crewmembers during spaceflight. It is important for exercise
to a maximum effort has been used to assess aerobic fitness prescriptions to be optimized to maintain astronauts’ fitness
and monitor and prescribe exercise training in athletes and to avoid premature physical and cognitive fatigue during
clinical populations with decades of controversy (Yeh et al., performance of high-risk, mission-critical tasks. Additionally,
1983; Poole et al., 2021). Noninvasive measurements of the more women are now flying into space, yet few studies about
ventilatory profile are represented commonly as two inflection sex-specific physiologic differences in the astronaut population
points: the ventilatory threshold (named as VT1 in this paper) have been explored during and after spaceflight. A reliable
and the respiratory compensation point (named as VT2 in and automated detection of VT1 and VT2 for prescribing exercise
this paper; Whipp et al., 1989; Cannon et al., 2009; Pallarés domains can help to quickly identify and guide individualized
et al., 2016; Carriere et al., 2019; Galán-Rioja et al., 2020). exercise prescriptions for the purpose of maintaining and
Determining VT1 and VT2 during an incremental exercise test enhancing crewmembers’ performance of submaximal
should be accompanied by a set of universally agreed upon extravehicular activities (EVAs) over long durations. We believe
and explicitly defined quality-control criteria (Meyer et al., that determining both VT1 and VT2 using the proposed novel
1996; Gaskill et al., 2001; Poole et al., 2016; Galán-Rioja et al., automated method and their associated absolute and relative
2020). Several VT methods for detecting VT have traditionally work rates can provide valuable information regarding
been based on a subjective visual analysis, but this is time- crewmembers’ ability to exercise (e.g., buffering capacity, lactate
consuming, requires 2–3 trained personnel with an independent kinetics) on the International Space Station (ISS) and to perform
reviewer and a strict set of criteria to maintain tight quality EVAs and lunar exploration. Understanding where VT1 and
control (Caiozzo et al., 1982; Yeh et al., 1983; Gladden et al., VT2 occur can assist in prescribing exercise intensity above
1985; Dickstein et al., 1990a; Shimizu et al., 1991). or below those points. Therefore, in this study, we determined
VT1 is commonly described as the point at which pulmonary the accuracy of our novel automated analysis by comparing
ventilation and carbon dioxide (CO2) output begin to increase the computerized results with subjective visual identification.
exponentially (Whipp et al., 1989; Dennis et al., 1992). The We also aimed to characterize absolute and relative work rates
ventilatory equivalent method has been used to identify VT1, at VT1 and VT2 and to identify if there are any differences
which is best described as the intensity of activity that causes in those values between female and male crewmembers.
the first rise in the ventilatory equivalent of oxygen (O2) without
a concurrent rise in the ventilatory equivalent of CO2 (Reinhard
et al., 1979; Powers et al., 1984). Determining VT1 using the MATERIALS AND METHODS
excess CO2 (ExCO2) method requires the intensity of exercise
that causes an increase from steady state to an excess production Subjects
of CO2 (Volkov et al., 1975). Likewise, the V-slope method, This study protocol was reviewed and approved by the NASA
also commonly used method, uses points that show an increase Johnson Space Center (JSC)‘s Institutional Review Board and
in the slope from less than 1 to greater than 1 in the CO2 in agreement with the Declaration of Helsinki. All participants
production (VCO2) by O2 consumption (VO2) data (Dickstein in the study provided signatures confirming their informed
et al., 1990a; Ekkekakis et al., 2008) and in the minute ventilation consent. Data were collected from 2013 to 2019. Astronaut
(VE) by VCO2 data (Beaver et al., 1986) for locating VT1 and participants in these data were active astronauts in flight training
VT2, respectively. All computerized methods mentioned require at JSC. Exact training status, e.g., amount of time spent in
finding the point where the slope changes markedly over the physically training, is not known. A total of 109 peak cycle
entire scattergram. However, two regression lines skim the tests were obtained from a large astronaut database (74 males
slope of data points from left to right and stop as soon as and 35 females) and used to examine the ventilatory thresholds
its criterion is satisfied; thus, the algorithm for searching VT1 at VT1 and VT2. The data set was separated into groups of
might estimate lower VO2 compared with the other methods female and males for the sex difference study (Pescatello et al.,
depending on the amount of noise in the data (Ekkekakis 2014). See Table 1 for mean participant characteristics and
et al., 2008). For the severe exercise intensity domain, VT2 is statistical significances for the overall and sex differences.
considered the second break point at which the partial pressure
of arterial CO2 starts to decline during heavy exercise (Beaver
et al., 1986; Nattie and Li, 2012). The 2-line regression model
for detecting the slope change in the scattergram also has TABLE 1 | Mean participant characteristics and statistical significances between
been used to find VT2 associated with VE and VCO2 (Volkov groups.
et al., 1975; Reinhard et al., 1979; Davis et al., 1980; Powers
Study Age (year) Mass (kg) Height (cm)
et al., 1984; Brooks, 1985; Beaver et al., 1986). However, a
disadvantage, inherent to methods that use regression analysis Overall 41.1 ± 7.3 76.8 ± 12.4 174.8 ± 9.4
during maximal exercise, is the possibility that the Male vs. Male 41.5 ± 6.8 82.6 ± 9.9 179.1 ± 5.8
hyperventilation phase (VT2 starting point) may be partially Female Female 40.3 ± 8.4 64.7 ± 7.2 165.7 ± 9.0
included in the calculation (Dickstein et al., 1990b). value of p 0.46 <0.0001 <0.0001

Frontiers in Physiology | www.frontiersin.org 2 December 2021 | Volume 12 | Article 782167

Kim et al. Automated VT and RCP Detection

Test Procedure and Data Collection was implemented as a MATLAB function, named findchangepts
Cardiorespiratory fitness (peak VO2) was determined by a (MATLAB® R2020a, The MathWorks, Inc., Natick, MA,
progressive, incremental graded cycle ergometer stress test to United States). Though the parametric global optimization
volitional exhaustion. Our graded exercise protocol was developed method identifies a data point change most significantly over
to measure peak VO2 and VT at JSC. The protocol has two the entire scattergram, it could be common during the
versions; a nominal protocol and a light protocol conducted determination of thresholds to find that some data are
on the LODE Excalibur sport cycle ergometer (Lode BV, indeterminate and inter-method differences are unavoidable in
Groningen, Netherlands). The nominal protocol consisted of nature (Gaskill et al., 2001). To provide a valid and reliable
a cycling warm-up at 50 W for 3 min followed by stepwise process, Gaskill et al. (2001) recommended to average the
increases of 25 W every minute until test termination. The combined multiple methods used to identify ventilatory threshold
light protocol consists of the same timed wattage increases (Gaskill et al., 2001). Thus, we identified VT1 using the mean
(i.e., 3 min warm-up then 1-min increases), but the wattage of ExCO2 and V-slope (Figure 1A) and detected VT2 using
starts at 45 W and with 15 W increases. Participants were the mean of the excess minute ventilation (ExVE) method
assigned the nominal or light protocol based on body weight. (Kim et al., 2020) and V-slope (Figure 1B). Figure 1C shows
Participants were assigned the light protocol if they weighed the combined VT1 and VT2 corresponded to gas exchange
less than 65 kilograms. Eight participants fell in this category. data: VE/VO2, ventilatory equivalent for O2 (blue dots); VE/
We observed that aerobic response was sufficiently fast to adjust VCO2, ventilatory equivalent for CO2 (red dots); PETO2, end-tidal
to the 15 or 25 W workload increases within the 1-min stage time.
Participants were instructed to maintain a cadence of 75
revolutions per minute (RPM) throughout testing. Respiratory A
gases were sampled per 10-s interval and analyzed using the
ParvoMedics TrueOne® 2400 metabolic cart. After a 30-min
warm-up, O2 and CO2 gas were calibrated using known gases
(16% O2, 4% CO2) and air flow was calibrated with a 3-L
syringe. To ensure accuracy of indirect calorimetry, gas and
flow calibration was conducted prior to every exercise test
and consisted of ambient and standard gas calibration along
with flow meter calibration. Calibration was accepted if new
calibration parameters were within +/− 3% of previous values.
Additionally, maintenance procedures were followed in B
accordance with manufacturer guidelines. The rating of perceived
exertion (RPE; Borg scale 6–20) was measured every 2–3 min
during the exercise test. Heart rate (HR) was determined from
12-lead electrocardiogram recordings throughout the test
(CardioSoft CASE®, GE Healthcare, WI, United States). The
test was considered to be maximal when at least 4 of 5 following
criteria were met: (1) a respiratory exchange ratio (RER) of
≥1.10, (2) a plateau in VO2 with increasing workloads, (3)
workload volitional fatigue (a fall of 10 RPM), (4) exercise
peak HR that was within 10 beats of the age-predicted maximal C
HR [207–(0.67 × age)] (Gellish et al., 2007), and (5) RPE at
or greater than 19. All participants reached at least 4 of the
listed criteria.
In summary, exercise variables in this study were work
rates (W), %Wmax (%), HR (beat/min), RER (VO2:VCO2),
and VE (L/min) at the ventilatory thresholds (VT1 and VT2)
and peak VO2 in absolute (L/min) and relative to body weight
(ml/kg/min) expressions. The first 3 min of warm-up were
excluded in data analysis.

Proposed Computerized Determination of FIGURE 1 | Example of determination of VT1 and VT2. (A) left: ExCO2, right:
VT Values V-slope for VT1. (B) left: ExVE, right: V-slope for VT2. (C) The combined VT1
Detection of abrupt change in data distribution has been and VT2 corresponded to gas exchange data: VE/VO2, ventilatory equivalent
considered as one of the important practical problems arising for O2 (blue dots); VE/VCO2, ventilatory equivalent for CO2 (red dots); PETO2,
end-tidal pressure of O2 (green dots); end-tidal pressure of CO2, PETCO2 (pink
in various applications (Riedel, 1994). In this study, we used
dots).
a parametric global optimization method (Lavielle, 2005), which

Frontiers in Physiology | www.frontiersin.org 3 December 2021 | Volume 12 | Article 782167

Kim et al. Automated VT and RCP Detection

FIGURE 2 | Code for the suggested approach.

A of metabolic reactions in the human body (Beaver et al., 1986).

At high intensity of exercise, the high-energy demand triggers
a breakdown of glucose and a reduction of pyruvate. These
metabolic processes produce lactate faster than the body can
metabolize it. Bicarbonate buffers hydrogen H+ as a
countermeasure when lactate flux increases faster than removal.
As exercise intensity increases above this point, VCO2 increases
and systemic pH in the bloodstream decreases as the H+ from
the bicarbonate is not sufficiently buffered (Myers and Ashley,
1997). This decrease in blood pH triggers carotid bodies,
chemoreceptor cells located in the carotid artery, to further
B
increase VE. Therefore, VT2 can be determined as the point
in time where VE increases to compensate for VCO2 being
greater than VO2, while end-tidal CO2 levels are reduced
(Pallarés et al., 2016). We reformulated the typical ExCO2
( )
concept [e.g., VCO 2 2 / VO 2 - VCO 2 )] (Volkov et al., 1975;
Gaskill et al., 2001) to the ExVE form
( )
such as VE 2 / VCO 2 - VE .
Figure 2 shows the code of the procedure how we generated
the input data, how we applied the findchangepks function to
detect each changepoint location, and how we finally determined
FIGURE 3 | Visual determination of VT1 (A) and VT2 (B). The arrows VT1 and VT2. For example, we specified “Statistic” as “std”
designate the choice of VT1 and VT2 in these sample graphs from one because the “std” option detects a significant change that occurs
subject. while the standard deviation of input data distribution increases
after ventilatory thresholds. Also, we specified “MaxNumChanges”
as “1” to return the index of the most significant change point
pressure of O2 (green dots); end-tidal pressure of CO2, PETCO2 in the scattergrams (see green dashed lines in Figures 1A,B).
(pink dots). Finally, we identified the work rate associated The algorithm for searching might be more robust and less
with VT1 and VT2 and compared absolute and relative influenced by unexpected peak noise at the initial or end time
sex difference. range. It has been reported that VT1 and VT2 commonly lie
at exercise intensities between 50 and 65% of VO2 and between
75 and 87% of VO2, respectively (Beaver et al., 1986; Pallarés
The ExVE is the method we proposed to identify the intensity et al., 2016; Cerezuela-Espejo et al., 2018; Galán-Rioja et al.,
of exercise that caused an increase from steady state to excess 2020). Thus, it is recommended to use a specific range of the
VE (Kim et al., 2020). Anaerobic exercise triggers a cascade input signals only, for example, trimming half of the signal

Frontiers in Physiology | www.frontiersin.org 4 December 2021 | Volume 12 | Article 782167

Kim et al. Automated VT and RCP Detection

length (i.e., from 30th percentile to 80th percentile for VT1, and ExCO2 methods), VT2 (average of V-slope for VT2 and
from 50th percentile to 100th percentile for VT2). ExVE methods), and peak, a 2-way factorial ANOVA was then
used to examine main effects and interactions (VT × Group)
for the sex difference study. If significance was found, the
Visually Evaluated VT Values appropriate Holm-Sidak multiple comparison post hoc test
For the validation study, three trained evaluators independently was performed.
and randomly evaluated the graphs of the data to determine Data were analyzed and figures generated using MATLAB
VT1 and VT2 values. For each determination, graphs were R2020a (The MathWorks, Inc., Natick, MA, United States) and
visually evaluated for the assessment of change in data GraphPad Prism (version 8.4.3, La Jolla, CA, United States)
distribution. Specifically, for VT1, evaluators assessed the with statistical significance set at p < 0.05. Equivalence testing
intensity of activity that causes the first sustained rise in the was conducted with the TOSTER package in the R
VE/VO2 without a concurrent rise in the ventilatory VE/ statistical software (R: A Language and Environment for Statistical
VCO2 (Figure 3A). The rise in VE/VO2 is in concurrence Computing, R Core Team, R Foundation for Statistical
with RER reaching 1.0. For VT2, evaluators assessed increase Computing, Vienna Austria, 2019).1 All data are reported as
in both the VE/VO2 and VE/VCO2 (Figure 3B). This rise in mean ± SD.
VE/VO2 and VE/VCO2 is in concurrence with a decrease in
PetCO2 (5). A detailed protocol to maintain tight quality
control over determination of VT1 and VT2 values was developed RESULTS
and included the following rules: (1) if, after concurrently
viewing all graphs, an evaluator still thought that the VT Subject Characteristics and Sex
value was indeterminate, then data for that subject were Comparison of Exercise Variables at VT1
rejected. If the evaluators thought that data were usable, they and VT2
then chose what they thought to be the most representative Males and females were matched for age but different for mass
value. (2) The values determined by the three independent and height (Table 1). Mean exercise variables for the overall
investigators were then compared by a fourth independent subjects and sex difference are reported in Table 2. At peak,
investigator. If the values determined by the evaluators were VT1, and VT2, females had lower absolute O2 uptake, W, and
within 1 exercise stage (50 W or less than 15%), then values ventilation and relative O2 uptake compared to men (p ≤ 0.04).
for the 3 investigators were averaged. (3) If values were within HRs and RERs were not different at peak exercise for either
15% of either of the initial investigators, then the VT values group. However, HRs and RERs were lower in the male group
were averaged. Comparison values greater than 15% were compared to the female group at VT1 and VT2 (p ≤ 0.02).
removed from the analysis. (4) The appearance time of VT
needed to be after the 75 W (after 4 min) warm-up of the
exercise test or the data were rejected and were considered
Relative Exercise Intensities at VT1 and VT2
Exercise intensities at VT1 and VT2 relative to peak VO2 and
to be indeterminate. The data reported for the visual
peak W for the combined analysis in all subjects and sex
identification method include only those participants whose
differences are reported in Figure 4. In the combined analysis,
data met all the criteria. Evaluators rejected 57 for VT1 and
VT1 occurred at 58% of peak VO2 and VT2 occurred at 79%
28 for VT2 based on our criteria. For the accurate comparison
of peak VO2 (p < 0.0001). Work rates at VT1 and VT2 were
between the visual analysis and the automated analysis, we only
50 and 69% of peak W (p < 0.0001). For the sex difference
used the paired matches.
comparison, VT1 and VT2 occurred at a greater relative percentage
of their peak VO2 for females (67 and 88%) compared to
males (55 and 74%), main effect for group; p < 0.001. However,
Statistical Analysis no differences were found for group expressed as a relative
Independent t tests were used when examining subject
percentage of peak W. Both sexes had VT1 at 50% and VT2
characteristics and sex comparison of exercise variables for
at 70% of peak W (p ≥ 0.40).
VO2, W, HR, RER, and VE values at VT1 and VT2.
The mean differences of the subjective analysis and the
automated analysis were examined using independent t test Comparison Between Visual and
to test for differences between measures. The Bland–Altman Automated
analysis (Giavarina, 2015) assessed the limits of agreement Independent t tests for the comparison between the subjective
between VT1 and VT2. Formal equivalence testing was conducted analysis and the automated analysis found no difference for
with the TOST approach with predefined equivalence bounds VT1 (p > 0.05) and VT2 (p > 0.05). Figure 5 illustrates the Bland–
of ±25 (Schuirmann, 1987). The data were also analyzed by Altman plot. Bias for VT1 was 7.08 ± 12.2 W with the 95%
the intraclass correlation coefficients (ICC) to examine the Limits of Agreement from −16.99 to 31.06 W. Bias for VT2
relationships between the subjective analysis and the was −2.2 ± 12.7 W with the 95% Limits of Agreement from
automated analysis. −27.0 to 22.7 W. For VT1 and VT2 combined, the equivalence
For each subject, absolute VO2, relative VO2, W, HR, RER,
and VE were determined at VT1 (average of V-slope for VT1 https://www.R-project.org
1

Frontiers in Physiology | www.frontiersin.org 5 December 2021 | Volume 12 | Article 782167

Kim et al. Automated VT and RCP Detection

TABLE 2 | Mean values at peak, VT1, and VT2.

VO2 (ml/kg/ RER

Study VO2 (L/min) Work rate (W) HR (beat/min) VE (L/min)
min) (VO2:VCO2)

Values at peak Overall 3.3 ± 0.9 43.2 ± 7.7 303.3 ± 76.4 177 ± 10 1.28 ± 0.08 140.0 ± 37.0
Male vs. Female Males 3.8 ± 0.6 45.8 ± 7.1 339.9 ± 48.4 178 ± 8 1.29 ± 0.07 156.4 ± 31.1
Females 2.4 ± 0.4 37.8 ± 5.8 225.9 ± 66.5 177 ± 13 1.28 ± 0.09 105.3 ± 21.2
value of P <0.0001 <0.0001 <0.0001 0.74 0.48 <0.0001
Values at VT1 Overall 1.9 ± 0.5 25.2 ± 4.5 156.2 ± 43.9 137 ± 15 0.99 ± 0.06 52.0 ± 11.3
Male vs. Female Males 2.1 ± 0.4 25.7 ± 4.8 172.3 ± 39.8 134 ± 14 1.00 ± 0.05 55.6 ± 10.4
Females 1.6 ± 0.3 24.0 ± 3.4 120.2 ± 28.9 144 ± 15 0.97 ± 0.08 44.5 ± 9.4
value of P <0.0001 0.04 <0.0001 <0.01 0.02 <0.0001
Values at VT2 Overall 2.6 ± 0.6 33.3 ± 6.3 213.8 ± 53.2 155 ± 17 1.10 ± 0.06 77.0 ± 15.3
Male vs. Female Males 2.8 ± 0.5 34.3 ± 6.3 234.3 ± 45.3 151 ± 17 1.10 ± 0.06 82.2 ± 13.7
Females 2.1 ± 0.4 31.3 ± 6.0 167.7 ± 39.3 162 ± 13 1.10 ± 0.08 65.8 ± 12.1
value of P <0.0001 0.01 <0.0001 <0.01 0.88 <0.0001

VO2, volume of oxygen; HR, heart rate; RER, respiratory exchange ratio; VE, ventilation; VT, ventilatory threshold.

FIGURE 4 | Exercise intensities at VT1 and VT2 relative to peak VO2 and peak W for the combined analysis (A) and sex differences (B). *Indicates group main effect
differences P < 0.001. ǂ indicates significant main effect difference between VT1 and VT2 p < 0.0001. Data reported as means ± SD.

test was significant, t(132) = −9.389, p < 0.0001, given equivalence the subjective analysis and the automated analysis. The ICC
bounds of −25 and 25 (on a raw scale) and an alpha of 0.05. between the subjective analysis and the automated analysis were
For VT1 and VT2 individually, the equivalence tests also were 0.821 for VT1 and 0.830 for VT2 with a 95% confidence interval.
significant, t(51) = −4.513, p < 0.0001 and t(80) = 5.812, p < 0.0001,
respectively, given equivalence bounds of −25 and 25 (on a
raw scale) and an alpha of 0.05. Equivalence testing showed DISCUSSION
that the automated and visual measures were statistically
equivalent (p < 0.0001) in all ways, each VT1 or VT2 and The advantages of a computerized method include faster,
combined. Finally, a high degree of reliability was found between objective, and automated data analysis as well as improvements

Frontiers in Physiology | www.frontiersin.org 6 December 2021 | Volume 12 | Article 782167

Kim et al. Automated VT and RCP Detection

A B

FIGURE 5 | Bland–Altman analysis between visual and automated method. (A) VT1, (B) VT2.

in reproducibility and repeatability. The aim of this study was participants’ steady state vs. progressive exercise test. When
to provide a novel, reliable, and computerized method for we examined the ventilatory points in relation to peak VO2,
automatically identifying VT1 and VT2. We demonstrated that the female group had greater relative VO2 associated with
our method was able to determine ventilatory thresholds VT1 (67%) and VT2 (88%). Beaver et al. (1986) reported
comparable to the visual analysis accomplished by 3 trained that VT1 occurred at 55% of peak VO2 and VT2 occurred
evaluators. We also determined the associated work rates at 75% of peak VO2 (Beaver et al., 1986) and a recent meta-
expressed as absolute and relative submaximal VO2 and W analysis reported that VT1 occurs at 50 to 60% of peak VO2
and reported that sex differences exist for VT1 and VT2. (Galán-Rioja et al., 2020). These values are similar to ours
Others have compared various computerized methods for as we report the whole sample was at 58 and 79% of peakVO2
determining VT1 and demonstrated that regression-based for VT1 and VT2, respectively. As noted earlier, these differences
methods provide considerably different results (Ekkekakis et al., in VT2 for the female group may be because of sensitivity
2008). For example, VT1 values detected by using 2 regression of the carotid body ventilatory drive caused by body
lines were significantly lower with weaker correlations compared temperature, blood osmolarity, pH, K+, H+ buffering by
to other computerized methods. Additionally, Pearson correlation bicarbonate, and the change in partial pressure of O2 (Galán-
coefficients in VO2 (L/min) between ExCO2 and V-slope for Rioja et al., 2020).
VT1 were 0.517 and 0.526 (i.e., a moderate positive relationship) It has been reported that microgravity affects females
in each sample group. In this study, Pearson correlation coefficient and males differently (Mark et al., 2014). It is important
in VO2 (L/min) detected by ExCO2 and V-slope for VT1 using to better understand these sex differences as the female
the parametric global optimization method was 0.81 (i.e., a representation in the astronaut corps is increasing, meaning
strong positive linear relationship), which is a stronger correlation more women will be eligible to fly in space than ever before.
coefficient compared to using 2 regression lines (Ekkekakis Under the microgravity environment, one of the sex-specific
et al., 2008). We also found that the strong positive relationship differences in exercise response is orthostatic intolerance
for Pearson correlation coefficient between computerized methods caused by plasma volume loss and cardiovascular adjustments
for determining VT2 (ExVE and V-slope for VT2) between (Goel et al., 2014). Females generally have smaller body
ExVE and V-slope for VT2 means that the parametric global size, lower absolute, and relative aerobic fitness and are
optimization method. In support of these, we found no difference weaker in upper and lower body strength, which have
between the subjective visual method by 3 trained evaluators implications for risk of fatigue and injury from muscle
and our automated method. Our novel and automated protocol strains during EVA and emergency egress (Harm et al.,
may increase the methodological consistency in both research 2001). Thus, understanding exercise countermeasures and
and clinical practice. the adaptations between sexes is of high relevance for the
It has been reported that maximal aerobic capacity is astronaut population. Our data are similar to others that
associated with VT1 (59 to 65%) and VT2 (84 to 87%) and report sex differences in gas exchange threshold for VT1.
that maximal lactate steady state corresponds to VT2 (Pallarés In this study, we also showed that VT2 differences
et al., 2016; Cerezuela-Espejo et al., 2018). We found that occurred between males and females. This may be because
when comparing to peak work rate, VT1 and VT2 were of differences in breathing adjustments to chemosensitivity,
associated with 50 and 70% of peak W, respectively. These thermoregulation, and menstrual cycle hormones (Beaver
differences may be because of age and fitness of our study et al., 1986; Kilbride et al., 2003; Hayashi et al., 2012).

Frontiers in Physiology | www.frontiersin.org 7 December 2021 | Volume 12 | Article 782167

Kim et al. Automated VT and RCP Detection

Others have suggested using exercise work rates above exercise on the ISS and to perform EVAs and lunar exploration.
and below the ventilatory breakpoints for VT1 and VT2 for Lastly, accurately tracking fitness pre-, in-, and post-flight
the prescription of exercise training to define exercise domains is of importance for guidance on the efficacy of exercise
such as moderate, heavy-and severe exercise intensity domains training prescriptions as countermeasures.
(Beaver et al., 1986; Nattie and Li, 2012). This gives an
individualized approach to prescribe exercise specific to the
metabolic demands. Previous exercise training DATA AVAILABILITY STATEMENT
countermeasures during spaceflight or analogues have used
a relative percentage of peak VO2 (e.g., continuous cycle The datasets presented in this article are not readily available
exercise for 30 min at 75% of peak VO2 and interval treadmill because the study dataset are not publicly available due to
sessions of 30 s to 4 min at nearly maximal intensity; Moore privacy laws and other restrictions. Requests to access the
Jr et al., 2014; Ploutz-Snyder et al., 2018). However, the datasets should be directed to KK, [email protected].
responses to these exercise prescriptions still have high
variability for maintaining fitness. For example, Moore et al.
reported that astronauts who have higher initial aerobic ETHICS STATEMENT
capacities are more prone to loss of cardiorespiratory fitness;
however, the reason is unknown and may be because of The studies involving human participants were reviewed and
the frequency, intensity, time, and progression of the exercise approved by NASA Johnson Space Center. The patients/participants
prescription (Moore et al., 2014). Our data suggest that sex provided their written informed consent to participate in this study.
should be considered when prescribing exercise
countermeasures and the prescriptions could be further
individualized by prescribing based on VT1, VT2, and AUTHOR CONTRIBUTIONS
peak VO2.
Notably, we acknowledge a limitation with our study. This KK: conceptualization, methodology, software, data curation, writing
includes not obtaining arterial lactate samples and blood gasses – original draft, writing – review and editing, and visualization.
to confirm the metabolic and ventilation breakpoints. Further ER: methodology, validation, formal analysis, data curation, writing
validation should include these measurements. – original draft, and writing – review and editing. BP and DF:
validation and writing – review and editing. MY: validation, formal
analysis, writing – original draft, and writing – review and editing.
CONCLUSION MD: investigation, resources, writing – review and editing,
supervision, project administration, and funding acquisition. All
In summary, the new automated method has been shown authors contributed to the article and approved the submitted
to identify inflection points in each of the variables used version.
to reliably determine VT1 and VT2. Furthermore, we show
that sex influences the VT1 and VT2 in members of the
US astronaut corps. Detection of both VT1 and VT2 and FUNDING
their associated absolute and relative work rates may provide
valuable information regarding crewmembers’ ability to This work was supported by the NASA Habitations System Account.

REFERENCES thresholds in runners: validity and reliability of exercise test performance

parameters. Front. Physiol. 9, 1–10. doi: 10.3389/fphys.2018.01320
Beaver, W. L., Wasserman, K., and Whipp, B. J. (1986). A new method for Davis, J. A., Whipp, B. J., and Wasserman, K. (1980). The relation of ventilation
detecting anaerobic threshold by gas exchange. J. Appl. Physiol. 60, 2020–2027. to metabolic rate during moderate exercise in man. Eur. J. Appl. Physiol.
doi: 10.1152/jappl.1986.60.6.2020 Occup. Physiol. 44, 97–108. doi: 10.1007/BF00421087
Brooks, G. A. (1985). Anaerobic threshold: review of the concept and directions Dennis, S. C., Noakes, T. D., and Bosch, A. N. (1992). Ventilation and blood lactate
for future research. Med. Sci. Sports Exerc. 17, 22–34. increase exponentially during incremental exercise. J. Sports Sci. 10, 437–449.
Caiozzo, V. J., Davis, J. A., Ellis, J. F., Azus, J. L., Vandagriff, R., Prietto, C., et al. Dickstein, K., Barvik, S., Aarsland, T., Snapinn, S., and Karlsson, J. (1990a).
(1982). A comparison of gas exchange indices used to detect the anaerobic A comparison of methodologies in detection of the anaerobic threshold.
threshold. J. Appl. Physiol. 53, 1184–1189. doi: 10.1152/jappl.1982.53.5.1184 Circulation 81, 38–46.
Cannon, D. T., Kolkhorst, F. W., and Buono, M. J. (2009). On the determination Dickstein, K., Barvik, S., Aarsland, T., Snapinn, S., and Millerhagen, J. (1990b).
of ventilatory threshold and respiratory compensation point via respiratory Validation of a computerized technique for detection of the gas exchange
frequency. Int. J. Sports Med. 30, 157–162. doi: 10.1055/s-0028- anaerobic threshold in cardiac disease. Am. J. Cardiol. 66, 1363–1367.
1104569 Ekkekakis, P., Lind, E., Hall, E. E., and Petruzzello, S. J. (2008). Do regression-
Carriere, C., Corrà, U., Piepoli, M., Bonomi, A., Merlo, M., Barbieri, S., et al. based computer algorithms for determining the ventilatory threshold agree?
(2019). Anaerobic threshold and respiratory compensation point identification J. Sports Sci. 26, 967–976. doi: 10.1080/02640410801910269
during cardiopulmonary exercise tests in chronic heart failure. Chest 156, Galán-Rioja, M. Á., González-Mohíno, F., Poole, D. C., and González-Ravé, J. M.
338–347. doi: 10.1016/j.chest.2019.03.013 (2020). Relative proximity of critical power and metabolic/Ventilatory
Cerezuela-Espejo, V., Courel-Ibáñez, J., Morán-Navarro, R., Martínez-Cava, A., thresholds: systematic review and meta-analysis. Sports Med., 1–13. doi:
and Pallarés, J. G. (2018). The relationship between lactate and ventilatory 10.1007/s40279-020-01314-8

Frontiers in Physiology | www.frontiersin.org 8 December 2021 | Volume 12 | Article 782167

Kim et al. Automated VT and RCP Detection

Gaskill, S. E., Ruby, B. C., Walker, A. J., Sanchez, O. A., Serfass, R. C., and Ploutz-Snyder, L. L., Downs, M., Goetchius, E., Crowell, B., English, K. L.,
Leon, A. S. (2001). Validity and reliability of combining three methods to Ploutz-Snyder, R., et al. (2018). Exercise training mitigates multisystem
determine ventilatory threshold. Med. Sci. Sports Exerc. 33, 1841–1848. doi: deconditioning during bed rest. Med. Sci. Sports Exerc. 50, 1920–1928. doi:
10.1097/00005768-200111000-00007 10.1249/MSS.0000000000001618
Gellish, R. L., Goslin, B. R., Olson, R. E., Mcdonald, A., Russi, G. D., and Poole, D. C., Burnley, M., Vanhatalo, A., Rossiter, H. B., and Jones, A. M.
Moudgil, V. K. (2007). Longitudinal modeling of the relationship between (2016). Critical power: an important fatigue threshold in exercise physiology.
age and maximal heart rate. Med. Sci. Sports Exerc. 39, 822–829. doi: 10.1097/ Med. Sci. Sports Exerc. 48, 2320–2334. doi: 10.1249/MSS.0000000000
mss.0b013e31803349c6 000939
Giavarina, D. (2015). Understanding bland altman analysis. Biochem. Med. 25, Poole, D. C., Rossiter, H. B., Brooks, G. A., and Gladden, L. B. (2021). The
141–151. doi: 10.11613/BM.2015.015 anaerobic threshold: 50+ years of controversy. J. Physiol. 599, 737–767. doi:
Gladden, L. B., Yates, J., Stremel, R. W., and Stamford, B. A. (1985). Gas 10.1113/JP279963
exchange and lactate anaerobic thresholds: inter and intraevaluator agreement. Powers, S. K., Dodd, S., and Garner, R. (1984). Precision of ventilatory and
J. Appl. Physiol. 58, 2082–2089. doi: 10.1152/jappl.1985.58.6.2082 gas exchange alterations as a predictor of the anaerobic threshold. Eur. J.
Goel, N., Bale, T. L., Epperson, C. N., Kornstein, S. G., Leon, G. R., Palinkas, L. A., Appl. Physiol. Occup. Physiol. 52, 173–177. doi: 10.1007/BF00433388
et al. (2014). Effects of sex and gender on adaptation to space: behavioral Reinhard, U., Müller, P., and Schmülling, R.-M. (1979). Determination of
health. J. Women's Health 23, 975–986. doi: 10.1089/jwh.2014.4911 anaerobic threshold by the ventilation equivalent in normal individuals.
Harm, D. L., Jennings, R. T., Meck, J. V., Powell, M. R., Putcha, L., Sams, C. P., Respiration 38, 36–42.
et al. (2001). Invited review: gender issues related to spaceflight: a NASA Riedel, K. S. (1994). Detection of Abrupt Changes: Theory and Application.
perspective. J. Appl. Physiol. 91, 2374–2383. doi: 10.1152/jappl.2001.91.5.2374 United Kingdom: Taylor&Francis.
Hayashi, K., Kawashima, T., and Suzuki, Y. (2012). Effect of menstrual cycle Schuirmann, D. J. (1987). A comparison of the two one-sided tests procedure
phase on the ventilatory response to rising body temperature during exercise. and the power approach for assessing the equivalence of average bioavailability.
J. Appl. Physiol. 113, 237–245. doi: 10.1152/japplphysiol.01199.2011 J. Pharmacokinet. Biopharm. 15, 657–680. doi: 10.1007/BF01068419
Kilbride, E., Mcloughlin, P., Gallagher, C. G., and Harty, H. R. (2003). Do Shimizu, M., Myers, J., Buchanan, N., Walsh, D., Kraemer, M., Mcauley, P.,
gender differences exist in the ventilatory response to progressive exercise et al. (1991). The ventilatory threshold: method, protocol, and evaluator
in males and females of average fitness? Eur. J. Appl. Physiol. 89, 595–602. agreement. Am. Heart J. 122, 509–516. doi: 10.1016/0002-8703(91)
doi: 10.1007/s00421-003-0853-z 91009-C
Kim, K. J., Buxton, R., Crowell, J., Downs, M., and Abercromby, A. (2020). Volkov, N. I., Shirkovets, E. A., and Borilkevich, V. E. (1975). Assessment of
An automated excess minute ventilation method to detect The respiratory aerobic and anaerobic capacity of athletes in treadmill running tests. Eur.
compensation. Med. Sci. Sports Exerc. 52, 37–39. doi: 10.1249/01. J. Appl. Physiol. Occup. Physiol. 34, 121–130. doi: 10.1007/BF0099
mss.0000670404.13236.08 9924
Lavielle, M. (2005). Using penalized contrasts for the change-point problem. Whipp, B. J., Davis, J. A., and Wasserman, K. (1989). Ventilatory control of
Signal Process. 85, 1501–1510. doi: 10.1016/j.sigpro.2005.01.012 the isocapnic buffering region in rapidly-incremental exercise. Respir. Physiol.
Mark, S., Scott, G. B., Donoviel, D. B., Leveton, L. B., Mahoney, E., Charles, J. B., 76, 357–367. doi: 10.1016/0034-5687(89)90076-5
et al. (2014). The impact of sex and gender on adaptation to space: executive Yeh, M. P., Gardner, R. M., Adams, T., Yanowitz, F., and Crapo, R. (1983).
summary. J. Women's Health 23, 941–947. doi: 10.1089/jwh.2014.4914 “anaerobic threshold”: problems of determination and validation. J. Appl.
Meyer, K., Hajric, R., Westbrook, S., Samek, L., Lehmann, M., Schwaibold, M., Physiol. 55, 1178–1186. doi: 10.1152/jappl.1983.55.4.1178
et al. (1996). Ventilatory and lactate threshold determinations in healthy
normals and cardiac patients: methodological problems. Eur. J. Appl. Physiol. Conflict of Interest: The authors declare that the research was conducted in
Occup. Physiol. 72, 387–393. doi: 10.1007/BF00242266 the absence of any commercial or financial relationships that could be construed
Moore, A. D. Jr., Downs, M. E., Lee, S. M., Feiveson, A. H., Knudsen, P., as a potential conflict of interest.
and Ploutz-Snyder, L. (2014). Peak exercise oxygen uptake during and
following long-duration spaceflight. J. Appl. Physiol. 117, 231–238. doi: Publisher’s Note: All claims expressed in this article are solely those of the
10.1152/japplphysiol.01251.2013 authors and do not necessarily represent those of their affiliated organizations,
Myers, J., and Ashley, E. (1997). Dangerous curves: a perspective on exercise, or those of the publisher, the editors and the reviewers. Any product that may
lactate, and the anaerobic threshold. Chest 111, 787–795. doi: 10.1378/chest.111.3.787 be evaluated in this article, or claim that may be made by its manufacturer, is
Nattie, E., and Li, A. (2012). Central chemoreceptors: locations and functions. not guaranteed or endorsed by the publisher.
Compr. Physiol. 2, 221–254. doi: 10.1002/cphy.c100083
Pallarés, J. G., Morán-Navarro, R., Ortega, J. F., Fernández-Elías, V. E., and Copyright © 2021 Kim, Rivas, Prejean, Frisco, Young and Downs. This is an
Mora-Rodriguez, R. (2016). Validity and reliability of ventilatory and blood open-access article distributed under the terms of the Creative Commons Attribution
lactate thresholds in well-trained cyclists. PLoS One 11, 1–16. doi: 10.1371/ License (CC BY). The use, distribution or reproduction in other forums is permitted,
journal.pone.0163389 provided the original author(s) and the copyright owner(s) are credited and that
Pescatello, L. S., Riebe, D., and Thompson, P. D. (2014). ACSM’s Guidelines the original publication in this journal is cited, in accordance with accepted academic
for Exercise Testing and Prescription. United States: Lippincott Williams practice. No use, distribution or reproduction is permitted which does not comply
and Wilkins. with these terms.

Frontiers in Physiology | www.frontiersin.org 9 December 2021 | Volume 12 | Article 782167