The Determination of Step Frequency in 3-min Incremental Step-in-Place Tests for Predicting Maximal Oxygen Uptake from Heart Rate Response in Taiwanese Adults

The maximal oxygen uptake (VO2max) prediction models established by step tests are often used for evaluating cardiorespiratory fitness (CRF). However, it is unclear which type of stepping frequency sequence is more suitable for the public to assess the CRF. Therefore, the main purpose of this study was to test the effectiveness of two 3-min incremental step-in-place (3MISP) tests (i.e., 3MISP30s and 3MISP60s) with the same total number of steps but different step-frequency sequences in predicting VO2max. In this cross-sectional study, a total of 200 healthy adults in Taiwan completed 3MISP30s and 3MISP60s tests, as well as cardiopulmonary exercise testing. The 3MISP30s and 3MISP60s models were established through multiple stepwise regression analysis by gender, age, percent body fat, and 3MISP-heart rate. The statistical analysis included Pearson’s correlations, the standard errors of estimate, the predicted residual error sum of squares, and the Bland–Altman plot to compare the measured VO2max values and those estimated. The results of the study showed that the exercise intensity of the 3MISP30s test was higher than that of the 3MISP60s test (% heart rate reserve (HRR) during 3MISP30s vs. %HRR during 3MISP60s = 81.00% vs. 76.81%, p < 0.001). Both the 3MISP30s model and the 3MISP60s model explained 64.4% of VO2max, and the standard errors of the estimates were 4.2043 and 4.2090 mL·kg−1·min−1, respectively. The cross-validation results also indicated that the measured VO2max values and those predicted by the 3MISP30s and 3MISP60s models were highly correlated (3MISP30s model: r = 0.804, 3MISP60s model: r = 0.807, both p < 0.001). There was no significant difference between the measured VO2max values and those predicted by the 3MISP30s and 3MISP60s models in the testing group (p > 0.05). The results of the study showed that when the 3MISP60s test was used, the exercise intensity was significantly reduced, but the predictive effectiveness of VO2max did not change. We concluded that the 3MISP60s test was physiologically less stressful than the 3MISP30s test, and it could be a better choice for CRF evaluation.


Introduction
According to a report from the World Health Organization (WHO), cardiovascular disease is the world's leading cause of death. The number of deaths from ischemic heart disease account for 16% of the total deaths in the world, and strokes account for 11% of the total deaths [1]. Cardiorespiratory fitness (CRF) is an important indicator of cardiovascular health [2][3][4]. Through CRF assessment, current and future health conditions can Int. J. Environ. Res. Public Health 2022, 19, 563 3 of 13 for 3-5 min, and the CRF level is finally evaluated according to the heart rate during or after exercise [7,24,27]. This stepping test requires high balance, coordination, and muscle strength of the lower extremities. It is not suitable for individuals with poor physical fitness or balance problems; extreme obesity; and lower extremity injuries [7,28,29]. Therefore, the choice of a submaximal exercise protocol for predicting VO 2max must meet the physical fitness levels of a variety of adults.
Recent cross-sectional studies have proposed another step test method, namely, the 3-min incremental step-in-place (3MISP) test, for predicting the VO 2max of healthy adults, and the results have been significant [30,31]. The 3MISP test movement is simple, requires less space and equipment, has a short duration, and does not require the professional operation of equipment. According to individual differences, the test uses the middle of the line between the anterior superior iliac spine and the midpoint of the patella as the target of knee elevation during stepping, and it requires no step-up box, so the safety index is higher than that of the step-up tests [31]. The VO 2max prediction model established by 3MISP can improve the accuracy of VO 2max prediction, and it has the advantages of indoor testing, convenient management, simplicity, and ease of use [31]. It can also be used for CRF tests of ordinary people in the homes, at work, or elsewhere. However, it is still unclear which type of 3MISP stepping frequency sequence is more suitable for CRF measurement by the general public. Different stepping frequencies may result in different physiological responses in participants. In the past, few studies have analyzed the effects of different stepping frequency sequences on individual physiological responses and the effectiveness of predicting VO 2max . Therefore, the purpose of this cross-sectional study was to evaluate the impact of two 3MISP tests with different step-frequency sequences (i.e., 3MISP 30s and 3MISP 60s tests) on the individual's physiological responses and the effectiveness of VO 2max estimation. In this study, we assumed that under the same total number of steps, the 3MISP 60s , with fewer changes in stepping frequency, would cause less physiological stress on participants and have a lower exercise intensity, and that the effectiveness of VO 2max estimation would not be inferior to that of the 3MISP 30s test, which has more changes in stepping frequency.

Study Design
This cross-sectional study was approved by the Institutional Review Board (IRB) of the Industrial Technology Research Institute (IRB No: IRB-APP-F02-106-009). The study protocol included two 3MISP tests (3MISP 30s and 3MISP 60s ) at a given step count (360 steps) with different frequency sequences, and a graded CPET. In this study, a Polar H10 Heart Rate Monitor (Polar Electro Oy, Finland) was used to continuously monitor the heart rate responses of the participants during the CPET and 3MISP tests. The VO 2max of the participants was directly measured by CPET to exhaustion on an electromagnetic bicycle ergometer (Excalibur Sport Ergometer, Lode BV, The Netherlands) and cardiopulmonary exercise testing system (Vmax Encore 29 System, VIASYS Healthcare Inc., Yorba Linda, CA, USA). Because there is a significant correlation between the directly measured VO 2max value and the heart rate change during the 3MISP tests [30,31], this study used 3MISP-heart rate (HR) as a predictor of VO 2max and combined age, gender, and physical characteristics to establish two VO 2max prediction models (the 3MISP 30s and 3MISP 60s models). To evaluate the reliability and validity of these two VO 2max prediction models, this study used the predicted residual error sum of squares (PRESS) cross-validation procedure to verify the models.

Subjects
In this cross-sectional study, 140 adults in Taiwan (i.e., the training group: 70 men, 70 women) were used for developing a VO 2max prediction model and another 60 Taiwanese adults (i.e., the testing group: 30 men, 30 women) were used for cross-verifying the prediction model. Table 1 shows the basic characteristics of the subjects in the training and testing groups. A total of 100 healthy men and 100 healthy women completed the test. Prior to the implementation of the study, all subjects completed the informed consent forms. In this study, the body mass index (BMI) was calculated by dividing the weight (kg) by the square of the height (meters). The participants' body weight and percent body fat (PBF) were measured by InBody ® 570 Body Composition Analyzer (Biospace, Inc., Seoul, Korea) [32]. All participants in this study were randomly and openly recruited, through social networks and posting advertisements in public spaces. Inclusion criteria were healthy adults aged 20-64 years, living in Taiwan. Pregnant women; individuals with lung, cardiovascular, or metabolic diseases; and those with neurological, muscular, or skeletal diseases were excluded.

3-min Incremental
Step-In-Place Tests In this study, all participants performed two 3MISP tests (i.e., the 3MISP 30s and 3MISP 60s tests) with the same total number of steps (360 steps) and different step-frequency sequences ( Figure 1). The 3MISP 30s test started at 80 SPM and then increased by 16 SPM every 30 s ( Figure 1A), in which there were 6 stages of step frequency. The 3MISP 60s test started at 96 SPM and then increased by 24 SPM every 60 s ( Figure 1B), in which there were 3 stages of step frequency. Before the 3MISP test, participants were asked to wear a chest strap-style Polar H10 heart rate monitor to continuously monitor their heart rate responses during stepping. Once the heart rate monitor was worn, the middle of the line connecting the participant's anterior superior iliac spine and the midpoint of the patella was measured as the target of their knee elevation when they stepped. After the start of the 3MISP test, participants needed to follow the rhythm of the metronome to raise their knees and step for 3 min while ensuring that the knees were raised to the required height each time. If the participant could not keep up with the rhythm or reach the required knee lift height for 30 s, the test was terminated and removed from the data analysis. In addition, for safety reasons, participants continued stepping at 80 SPM for 30 s after the 3MISP test and then completed 30 s of standing rest. The heart rates of each participant at the beginning (HR0) of the 3MISP exercise, as well as at the 1st (HR1), 2nd (HR2), and 3rd (HR3) minutes of exercise, and at the 1st minute (HR4) after the end of the 3MISP test, were recorded for subsequent data analysis.

Cardiopulmonary Exercise Testing
In this study, the VO2max of all participants was measured directly on the electromagnetic bicycle ergometer with graded CPET until volitional exhaustion and with the cardiopulmonary exercise testing system. During the CPET process, participants wore a Polar H10 heart rate monitor and a suitable air-gathering mask (Hans-Rudolph), and the Borg Rating of Perceived Exertion (RPE, 6-20 scale) was used to evaluate the participants' degree of fatigue during the test. The heart rate monitor was mainly used to monitor the heart rate responses of the participants during the CPET process. The breath-by-breath air volume and the contents of O2 and CO2 were measured through the sampling line connected to the mask and the digital flow sensor. The initial load of CPET was 25 W, and then the resistance was increased by 15 W every 2 min until the participant could no longer maintain the pedaling rate of 70 revolutions per minute. In this study, the criteria for VO2max was the participant meeting three of the following conditions: a respiratory ratio (CO2/O2) ≥ 1.10, a maximum heart rate of more than 90% of the age-predicted maximal HR (220-age), and an RPE > 17; even if the exercise load was increased, the oxygen uptake still tended to be stable [7,31,33].

Statistical Analysis
In this study, the statistical software SPSS (version 20, IBM Corp., New York, NY, USA) was used for statistical analysis. All values are presented in the form of means ± standard deviations. The significance level was set to p < 0.05. The normality assumption was checked using Shapiro-Wilk test. The independent samples t-test (normally distributed data) and Mann-Whitney U test (non-normally distributed data) were used to compare the physical characteristics and heart rate responses between the training and testing groups. Cohen's d effect sizes were calculated to compare the practical difference between groups, where the effect sizes < 0.2, 0.2~0.5, 0.5~0.8, and >0.8 were indictive of a trivial, small, moderate, and large difference, respectively [34]. Pearson's correlation coefficients were used to analyze the linear relationship between the actual measured value of VO2max and gender, age, PBF, and 3MISP-HR in the training group, and the validity of the VO2max prediction model. The correlation coefficients were interpreted by Schober et al. (2018), with 0.90~1.00 categorized as very strong, 0.70~0.89 as strong, 0.40~0.69 as moderate, and 0.10~0.39 as weak correlation [35]. Through multiple stepwise regression analysis, two VO2max prediction models (i.e., the 3MISP30s and 3MISP60s models) were established by five parameters: gender, age, PBF, HR0, and ΔHR3−HR4. The multiple coefficients of determination (R 2 ), the absolute standard errors of estimate (SEE), and relative SEE (SEE%) were used to evaluate and compare the accuracies of the two VO2max prediction models. The PRESS statistical method was used to cross-validate the VO2max prediction model [31,36].

Cardiopulmonary Exercise Testing
In this study, the VO 2max of all participants was measured directly on the electromagnetic bicycle ergometer with graded CPET until volitional exhaustion and with the cardiopulmonary exercise testing system. During the CPET process, participants wore a Polar H10 heart rate monitor and a suitable air-gathering mask (Hans-Rudolph), and the Borg Rating of Perceived Exertion (RPE, 6-20 scale) was used to evaluate the participants' degree of fatigue during the test. The heart rate monitor was mainly used to monitor the heart rate responses of the participants during the CPET process. The breath-by-breath air volume and the contents of O 2 and CO 2 were measured through the sampling line connected to the mask and the digital flow sensor. The initial load of CPET was 25 W, and then the resistance was increased by 15 W every 2 min until the participant could no longer maintain the pedaling rate of 70 revolutions per minute. In this study, the criteria for VO 2max was the participant meeting three of the following conditions: a respiratory ratio (CO 2 /O 2 ) ≥ 1.10, a maximum heart rate of more than 90% of the age-predicted maximal HR (220-age), and an RPE > 17; even if the exercise load was increased, the oxygen uptake still tended to be stable [7,31,33].

Statistical Analysis
In this study, the statistical software SPSS (version 20, IBM Corp., New York, NY, USA) was used for statistical analysis. All values are presented in the form of means ± standard deviations. The significance level was set to p < 0.05. The normality assumption was checked using Shapiro-Wilk test. The independent samples t-test (normally distributed data) and Mann-Whitney U test (non-normally distributed data) were used to compare the physical characteristics and heart rate responses between the training and testing groups. Cohen's d effect sizes were calculated to compare the practical difference between groups, where the effect sizes <0.2, 0.2~0.5, 0.5~0.8, and >0.8 were indictive of a trivial, small, moderate, and large difference, respectively [34]. Pearson's correlation coefficients were used to analyze the linear relationship between the actual measured value of VO 2max and gender, age, PBF, and 3MISP-HR in the training group, and the validity of the VO 2max prediction model. The correlation coefficients were interpreted by Schober et al. (2018), with 0.90~1.00 categorized as very strong, 0.70~0.89 as strong, 0.40~0.69 as moderate, and 0.10~0.39 as weak correlation [35]. Through multiple stepwise regression analysis, two VO 2max prediction models (i.e., the 3MISP 30s and 3MISP 60s models) were established by five parameters: gender, age, PBF, HR0, and ∆HR3−HR4. The multiple coefficients of determination (R 2 ), the absolute standard errors of estimate (SEE), and relative SEE (SEE%) were used to evaluate and compare the accuracies of the two VO 2max prediction models. The PRESS statistical method was used to cross-validate the VO 2max prediction model [31,36]. The PRESS R 2 (R 2 p ) is defined as 1-(PRESS/SS total ), while the PRESS SEE (SEE p ) is calculated as √ RRESS/N [37]. Bland-Altman plots were used to compare the differences between the actual measured and estimated values of VO 2max , which allowed us to visualize the data distribution around the line of zero [17,23,38]. Table 2 lists the heart rate responses of the training group and the testing group during the 3MISP 30s and 3MISP 60s tests. There was no significant difference in the heart rate responses during the 3MISP 30s or 3MISP 60s test between the two groups (p > 0.05). Figure 2 shows the heart rate responses of all participants in these two 3MISP tests, and their corresponding exercise intensities. The results of the study showed that the heart rate of the participants at the third minute during the 3MISP 30s test was significantly higher than the heart rate at the third minute during the 3MISP 60s test (3MISP 30s : 156 ± 14 bpm, 3MISP 60s : 151 ± 15 bpm, p < 0.001). The exercise intensity tested by 3MISP 30s was higher than the exercise intensity tested by 3MISP 60s (%heart rate reserve (HRR) during 3MISP 30s vs. %HRR during 3MISP 60s = 81.00% vs. 76.81%, p < 0.001). Figure 3 shows the correlations between the actual measured VO 2max and age, PBF, and 3MISP-HR in the training group. Pearson's correlation coefficients showed a positive correlation between gender (women = 0, men = 1) and VO 2max (r = 0.453, p < 0.001). There was a negative correlation between VO 2max and age or PBF (age: r = −0.324, PBF: r = −0.671, both p < 0.001). There was a negative correlation between HR0 and VO 2max (3MISP 30s : r = −0.501, 3MISP 60s : r = −0.501, both p < 0.001). There was a positive correlation between ∆HR3−HR4 and VO 2max (3MISP 30s : r = 0.620, 3MISP 60s : r = 0.564, both p < 0.001).

Results
Two multiple regression models (i.e., the 3MISP 30s and 3MISP 60s models) for predicting VO 2max using age, gender, PBF, HR0, and ∆HR3−HR4 are presented in Table 3, as well as the results of cross-validation. All variables in the 3MISP 30s and 3MISP 60s models were independently related to VO 2max . In this study, the two VO 2max prediction models showed similar multiple correlations and SEE (3MISP 30s model: R 2 = 0.644, SEE = 4.2043 mL·kg −1 ·min −1 ; 3MISP 60s model: R 2 = 0.644, SEE = 4.2090 mL·kg −1 ·min −1 ). The cross-validation result of the PRESS statistical method showed that the changes in R 2 and SEE of the 3MISP 30s and 3MISP 60s models were very small (∆R 2 ≤ 0.007, ∆SEE ≤ 0.075 mL·kg −1 ·min −1 ).   Figure 3 shows the correlations between the actual measured VO2max and age, PBF, and 3MISP-HR in the training group. Pearson's correlation coefficients showed a positive correlation between gender (women = 0, men = 1) and VO2max (r = 0.453, p < 0.001). There was a negative correlation between VO2max and age or PBF (age: r = −0.324, PBF: r = −0.671, both p < 0.001). There was a negative correlation between HR0 and VO2max (3MISP30s: r = −0.501, 3MISP60s: r = −0.501, both p < 0.001). There was a positive correlation between ΔHR3−HR4 and VO2max (3MISP30s: r = 0.620, 3MISP60s: r = 0.564, both p < 0.001). , heart rate at the beginning of the 3MISP test; ΔHR3−HR4, the difference between the heart rate at the 3rd minute during the exercise and the heart rate at the 1st minute after the end of the 3MISP test.
Two multiple regression models (i.e., the 3MISP30s and 3MISP60s models) for predicting VO2max using age, gender, PBF, HR0, and ∆HR3−HR4 are presented in Table 3, as well as the results of cross-validation. All variables in the 3MISP30s and 3MISP60s models were independently related to VO2max. In this study, the two VO2max prediction models showed similar multiple correlations and SEE (3MISP30s model: R 2    (B) %HRR (the percentage of heart rate reserve) curves under the 3MISP30s and 3MISP60s tests; *** Significant difference (p < 0.001) in heart rate (A) or %HRR (B) between the 3MISP30s and 3MISP60s tests. Figure 3 shows the correlations between the actual measured VO2max and age, PBF, and 3MISP-HR in the training group. Pearson's correlation coefficients showed a positive correlation between gender (women = 0, men = 1) and VO2max (r = 0.453, p < 0.001). There was a negative correlation between VO2max and age or PBF (age: r = −0.324, PBF: r = −0.671, both p < 0.001). There was a negative correlation between HR0 and VO2max (3MISP30s: r = −0.501, 3MISP60s: r = −0.501, both p < 0.001). There was a positive correlation between ΔHR3−HR4 and VO2max (3MISP30s: r = 0.620, 3MISP60s: r = 0.564, both p < 0.001). , heart rate at the beginning of the 3MISP test; ΔHR3−HR4, the difference between the heart rate at the 3rd minute during the exercise and the heart rate at the 1st minute after the end of the 3MISP test.

Measured VO 2max (ml·kg
Two multiple regression models (i.e., the 3MISP30s and 3MISP60s models) for predicting VO2max using age, gender, PBF, HR0, and ∆HR3−HR4 are presented in Table 3, as well as the results of cross-validation. All variables in the 3MISP30s and 3MISP60s models were independently related to VO2max. In this study, the two VO2max prediction models showed similar multiple correlations and SEE (3MISP30s model: R 2 (N = 140), showing the regression line. HR0, heart rate at the beginning of the 3MISP test; ∆HR3−HR4, the difference between the heart rate at the 3rd minute during the exercise and the heart rate at the 1st minute after the end of the 3MISP test.  Figure 4 shows the Pearson's correlations in the testing group between the measured VO 2max values and those predicted by the 3MISP 30s and 3MISP 60s models. The results of this study showed that the VO 2max values estimated by both models were highly correlated with the actual measured values of VO 2max (3MISP 30s : r = 0.804, 95% confidence interval (CI) = 0.715-0.872; 3MISP 60s model: r = 0.807, 95% CI = 0.720-0.875; both p < 0.001). The prediction values (R 2 ) from the 3MISP 30s and 3MISP 60s models were moderate, and the prediction errors were relatively large (3MISP 30s : R 2 p = 0.646, SEE p = 4.180 mL·kg −1 ·min −1 , SEE p % = 12.31%; 3MISP 60s : R 2 p = 0.651, SEE p = 4.134 mL·kg −1 ·min −1 , and SEE p % = 12.18%). ence between the heart rate at the 3rd minute during the exercise and the heart rate at the 1st minute after the end of the 3MISP test; R 2 p, PRESS squared multiple correlation coefficient; SEE, standard error of estimate; SEE%, SEE/mean of measured VO2max × 100; and SEEp, PRESS SEE. Figure 4 shows the Pearson's correlations in the testing group between the measure VO2max values and those predicted by the 3MISP30s and 3MISP60s models. The results of th study showed that the VO2max values estimated by both models were highly correlate with the actual measured values of VO2max (3MISP30s: r = 0.804, 95% confidence interv (CI) = 0.715-0.872; 3MISP60s model: r = 0.807, 95% CI = 0.720-0.875; both p < 0.001). Th prediction values (R 2 ) from the 3MISP30s and 3MISP60s models were moderate, and th prediction errors were relatively large (3MISP30s: R 2 p = 0.646, SEEp = 4.180 mL·kg −1 ·min SEEp% = 12.31%; 3MISP60s: R 2 p = 0.651, SEE p = 4.134 mL·kg −1 ·min −1 , and SEEp% = 12.18%)   Figure 5 shows the differences between the predicted VO 2max values and those measured by the 3MISP 30s and 3MISP 60s models with the 95% limits of agreement (LoAs) in Bland-Altman Plots. The average differences between the measured VO 2max values and those predicted by the 3MISP 30s model (LoAs = −8.60 to 7.86 mL·kg −1 ·min −1 ) and 3MISP 60s model (LoAs = −8.16 to 8.19 mL·kg −1 ·min −1 ) were −0.37 and 0.02 mL·kg −1 ·min −1 , respectively, and none of them reached the level of significance (p > 0.05). However, the 3MISP 30s and 3MISP 60s models seem to underestimate the real VO 2max of participants with high VO 2max values, whereas they overestimate it for participants with low VO 2max values. Health 2022, 19, x FOR PEER REVIEW 9 of 13 Figure 5 shows the differences between the predicted VO2max values and those measured by the 3MISP30s and 3MISP60s models with the 95% limits of agreement (LoAs) in Bland-Altman Plots. The average differences between the measured VO2max values and those predicted by the 3MISP30s model (LoAs = −8.60 to 7.86 mL·kg −1 ·min −1 ) and 3MISP60s model (LoAs = −8.16 to 8.19 mL·kg −1 ·min −1 ) were −0.37 and 0.02 mL·kg −1 ·min −1 , respectively, and none of them reached the level of significance (p > 0.05). However, the 3MISP30s and 3MISP60s models seem to underestimate the real VO2max of participants with high VO2max values, whereas they overestimate it for participants with low VO2max values.

Discussion
CRF is one of the basic elements of healthy physical fitness. For the assessment of large-scale CRF, the subjects have a large age span and different physical conditions. To determine which step frequency sequence is more appropriate for healthy adults to use in

Discussion
CRF is one of the basic elements of healthy physical fitness. For the assessment of large-scale CRF, the subjects have a large age span and different physical conditions. To determine which step frequency sequence is more appropriate for healthy adults to use in CRF evaluation, this cross-sectional study compared the heart rate responses in 3MISP 30s and 3MISP 60s tests and the accuracies of VO 2max prediction models. The results of this research show that the 3MISP 60s test, which has fewer step-frequency changes, significantly reduces the exercise intensity while maintaining the same total number of steps (work amount). The subject's heart rate responses in the 3MISP 60s test are lower than those in the 3MISP 30s test. We also confirmed that the predictive effectiveness of VO 2max did not decrease when the exercise intensity was reduced. Therefore, a simple, safe, and effective 3MISP 60s test program may be a better choice as a field test for evaluating CRF across age groups.
During the progressive stepping test, the ability of the subject to perform incremental exercise loads, and the physiological response to exercise loads, can be observed [39]. The heart rate is an important physiological indicator of the function of the heart and the circulatory system. The heart rate, measured continuously, can simply, objectively, and directly reflect exercise intensity and the individual's adaptability to the exercise load, without the requirement of familiarization trials [3,16]. Therefore, this study analyzed the individual heart rate responses during the 3MISP 30s and 3MISP 60s tests using continuous heart monitoring at different step-frequency sequences. The results of this study show that under the condition of the same total number of steps, the 3MISP 60s test, with fewer step-frequency changes, significantly reduces the exercise intensity at the second and third minutes ( Figure 2B). The heart rate changes (HR0 and ∆HR3−HR4) during the 3MISP 30s and 3MISP 60s tests were all significantly correlated with the actual measured VO 2max (Figure 3B,C). This is consistent with the results of other studies using the steptest scheme to predict VO 2max and suggests that the CRF test, the subject's basal heart rate, recovery heart rate, and heart rate during exercise are all negatively correlated with measured VO 2max [21,29,30,36,40]. Obviously, the change of heart rate during the 3MISP 30s and 3MISP 60s tests is one of the important factors for estimating VO 2max . In the case of maintaining the same total number of steps, decreasing the times of 3MISP step-frequency changes can reduce the exercise intensity as well as improve safety. Because the subjects bear relatively low physiological load during the 3MISP 60s test, it may help to encourage more people to complete the CRF test. The relatively low physiological load can also increase the completion of the test movements and improve the effectiveness of the VO 2max prediction formula.
In this study, age, gender, and PBF were significantly correlated with measured VO 2max ( Figure 3A), which is consistent with previous research results. In the past, many studies have shown that age, gender, and physical characteristics (BMI/PBF/waist circumference (WC)) are important factors in estimating VO 2max . The addition of characteristic variables to the VO 2max prediction models led to decreased error, which may have been caused by the improvement of individual's characteristics in personalization [41]. The PBF's prediction of VO 2max is more accurate than those of BMI and WC, and it is a key factor for estimating VO 2max [21,30,36,42,43]. Therefore, in this study, age, gender, and PBF were used to establish the VO 2max estimation formula. In addition, heart rate is one of the variables most commonly used to develop the predictive models [16,41]. To further improve the accuracy of VO 2max estimation, this cross-sectional study also used the heart rate changes during the 3MISP 30s and 3MISP 60s tests as the predictive variable to establish the VO 2max estimation formula. The two VO 2max estimated models (i.e., the 3MISP 30s and 3MISP 60s models) based on age, gender, PBF, HR0, and ∆HR3−HR4 can explain 64.4% of the measured VO 2max . The SEE values of the two models were also very close (Table 3). This shows that the 3MISP 60s test, despite its lower exercise intensity, is similar to the 3MISP 30s test. It can also predict the VO 2max of healthy adults very well (R 2 = 0.644, SEE = 4.209 mL·kg −1 ·min −1 ). It can be seen that the prediction accuracy of VO 2max for submaximal exercises with lower intensity is not worse than that for submaximal exercises with higher exercise intensity.
In past cross-sectional studies, step-up tests have often been used to estimate adult VO 2max [20,21,29,36,39], with considerable results. According to age, gender, body weight, height, BMI, and heart rate during step-up tests, the R 2 and SEE of the VO 2max estimation formula are 0.56-0.73 and 4.05-5.01 mL·kg −1 ·min −1 , respectively. However, step-up tests usually require a step box with a height of 20-50 cm. Participants who are overweight or have knee injuries, abnormal gait, or impaired balance are likely to fall while stepping onto and off the box [30]. In the step-in-place test, the middle of the line between the anterior superior iliac spine and the midpoint of the patella is used as the target of knee elevation during stepping. No step-up box is required, so it is safer than the step-up test [30,31,44]. At present, the step-in-place test is mainly a part of the senior fitness test, which is used to assess the aerobic endurance of older adults (≥60 years). However, few related studies have examined the measurement of the CRF level of younger adults by the step-in-place test. To further improve the safety and universality of adult CRF examination, this study designed two 3MISP test methods (3MISP 30s and 3MISP 60s ) with different step-frequency sequences. Compared with previous studies that used the step-up test to predict VO 2max , the R 2 (0.644) and SEE (4.2043-4.2090 mL·kg −1 ·min −1 ) values of the two VO 2max prediction models established using 3MISP 30s and 3MISP 60s tests in this study are both acceptable (Table 3). They can predict the VO 2max of healthy adults relatively accurately.
The cross-validation results of this study showed that both the 3MISP 30s and 3MISP 60s models had a high level of cross-validity ( Table 3). The measured VO 2max values and those predicted by the 3MISP 30s and 3MISP 60s models were highly correlated in the testing group ( Figure 4). From Bland-Altman Plots, it was also found that the measured and estimated VO 2max obtained from the 3MISP 30s and 3MISP 60s models had high consistency ( Figure 5). The accuracy of 3MISP-based predictions in this cross-sectional study was consistent with the previous research reporting that the VO 2max models were moderately to highly accurate [16]. There were significant relationships and agreements between the measured VO 2max and those estimated by submaximal exercise-based prediction models [16,17]. These results provided evidence for the effectiveness of the 3MISP 30s and 3MISP 60s models in predicting VO 2max . Therefore, the two 3MISP tests with different step-frequency sequences proposed in this study, and the established VO 2max prediction models, are practicable. However, the exercise intensity of the 3MISP 60s test is relatively low, and the physiological load on the subjects is relatively small. Therefore, it may be a better choice to measure the adult CRF level in comparison with the 3MISP 30s test.
In summary, this study has three features. First, the two self-designed step-in-place test methods have the advantages of simplicity and ease of use. Second, under the condition of the same total number of steps (work amount), the 3MISP 60s test method, with fewer step-frequency changes, significantly reduces the exercise intensity and improves safety. Finally, it was successfully verified that the predictive effectiveness of VO 2max did not decrease when the exercise intensity was reduced.
There are several limitations in this cross-sectional study. A limitation of the study is that the mean age of all subjects was 43.72 ± 10.09 years, who are representative of healthy adults living in Taiwan. Thus, we could not verify whether the results may apply to older individuals, patients, or other racial groups. Another potential limitation is that our study established VO 2max prediction models using the cycle ergometer, a popular exercise mode, rather than the treadmill ergometry. Finally, although we have developed two 3MISP tests with different step-frequency sequences, the optimal step frequency is not known in this study, and further research will be needed to determine it.

Conclusions
The 3MISP 30s and 3MISP 60s tests are simple to operate, take up little time and space, and do not require expensive equipment or professionals to operate said equipment. This study has confirmed that the 3MISP tests of these two different step-frequency sequences can effectively predict the VO 2max of healthy adults. The 3MISP 60s test, despite its lower exercise intensity, is not inferior to the 3MISP 60s test in terms of the accuracy of predicting VO 2max . Moreover, the 3MISP 60s test places a smaller physiological burden on the subjects than that of the 3MISP 30s test, and it is safer for CRF field tests in adult populations, and especially in older subjects. In this study, the 3MISP 60s test may be a better choice than the 3MISP 30s test for measuring the CRF level of healthy adults. However, a safer and more effective CRF test method for submaximal exercise is still to be pursued by researchers.