Detection of ventilatory thresholds using near-infrared spectroscopy with a polynomial regression model.

Whether near-infrared spectroscopy (NIRS) is a convenient and accurate method of determining first and second ventilatory thresholds (VT1 and VT2) using raw data remains unknown. This study investigated the reliability and validity of VT1 and VT2 determined by NIRS skeletal muscle hemodynamic raw data via a polynomial regression model. A total of 100 male students were recruited and performed maximal cycling exercises while their cardiopulmonary and NIRS muscle hemodynamic data were measured. The criterion validity of VT1VET and VT2VET were determined using a traditional V-slope and ventilatory efficiency. Statistical significance was set at α = . 05. There was high reproducibility of VT1NIRS and VT2NIRS determined by a NIRS polynomial regression model during exercise (VT1NIRS, r = 0.94; VT2NIRS, r = 0.93). There were high correlations of VT1VET vs VT1NIRS (r = 0.93, p < .05) and VT2VET vs VT2NIRS (r = 0.94, p < .05). The oxygen consumption (VO2) between VT1VET and VT1NIRS or VT2VET and VT2NIRS was not significantly different. NIRS raw data are reliable and valid for determining VT1 and VT2 in healthy males using a polynomial regression model. Skeletal muscle raw oxygenation and deoxygenation status reflects more realistic causes and timing of VT1 and VT2.

Introduction

Maximal oxygen consumption (VO2max) is considered the gold standard for evaluating cardiorespiratory function (Mitchell et al., 1958, Saltin and Astrand, 1967). The submaximal anaerobic threshold (AT) and ventilatory threshold (VT) provide a better clinical index of aerobic fitness than VO2max. Previous studies confirmed that the VT was also related to the intensity of aerobic exercise, self-reported activity levels, and training benefits (Tamai et al., 1993, Gaskill et al., 2001). Therefore, developing a convenient and accurate method to determine VT is a significant research issue.

Other methods of estimating the VT have been developed such as tissue deoxygenation (Bhambhani et al., 1997), double product breakpoint (Omiya et al., 2004), non-linear increased EMG activity (Lucia et al., 1997), heart rate deflection (Conconi et al., 1982), and heart rate variability (Cottin et al., 2007). One study indicated that changes in the respiratory frequency were related to the first ventilatory threshold (VT1) and second ventilatory threshold (VT2) (Neder and Stein, 2006). The most popular technique is the V-slope method (Beaver et al., 1986), which determines the VT by measuring the non-linear point of increase in the slope of carbon dioxide production (VCO2) vs oxygen uptake (VO2) during incremental exercise. The VT reflects lactic acid accumulation during exercise that causes hyperventilation. (Myers and Ashley, 1997).

Near-infrared spectroscopy (NIRS) signals obtained during exercise were confirmed to reflect local tissue oxygen delivery, utilization, and blood flow (Boushel et al., 2001, DeLorey et al., 2003). The primary NIRS parameters include oxygenated hemoglobin (O2Hb), oxygenated myoglobin (O2Mb), deoxygenated hemoglobin (HHb), and deoxygenated myoglobin (HMb). O2Hb and O2Mb were confirmed to reflect tissue O2 delivery. HHb and HMb were confirmed to reflect tissue O2 utilization.

The physiological significance of muscle oxygen saturation was demonstrated using different exercise intensities and modes. Previous studies reported that skeletal muscle oxygenation was relatively constant during walking and gradually increased during incremental running (Hiroyuki et al., 2002, Lee et al., 2011). Other studies indicated that muscle oxygenation status follows a sigmoidal profile during incremental cycling (Ferreira et al., 2007, Legrand et al., 2007, Thomas and Stephane, 2008, Racinais et al., 2014, Boone et al., 2015). Studies reported that muscle O2 delivery (O2Hb + O2Mb) increased until moderate intensity and then decreased or was maintained until exercise ended. Other studies found that O2 delivery decreased from the start to the end of exercise. Further studies indicated that muscle O2 utilization (HHb + HMb) increased from the beginning to the end of exercise. Other studies found that O2 utilization increased to high intensity and were then maintained until the end of exercise.

Previous studies indicated that VT1 and VT2 could be determined using NIRS based on the breakpoint of skeletal muscle oxygenation and blood flow (deoxy[Hb + Mb] and total hemoglobin [Hb + Mb]) (Boone et al., 2015), muscle oxygenation (O2Hb) (Legrand et al., 2007, Racinais et al., 2014), the difference between muscle oxygenation and deoxygenation (Δ[O2HbMb-HHbMb]) (van der Zwaard et al., 2016), and respiratory muscle oxygenation (Moalla et al., 2005). Mathematical models of determining VT1 and VT2 include single linear regression (Legrand et al., 2007), double linear regression (Grassi et al., 1999, Moalla et al., 2005, Legrand et al., 2007, Wang et al., 2012), non-linear models (Racinais et al., 2014), and combining linear and polynomial regression (van der Zwaard et al., 2016). Integrating previous studies, the choice of parameters for determining VT1 and VT2 were muscle oxygenation (O2Hb + O2Mb) or deoxygenation (HHb + HMb). The muscle oxygenation and deoxygenation data in most mathematical models are polynomial patterns (sigmoidal profile).

Although NIRS was confirmed as a non-invasive method to reflect oxygen delivery, utilization, and blood flow during exercise, whether oxygenation and deoxygenation NIRS signals can be used to determine VT1 and VT2 requires further clarification. Establishing mathematical models using raw data can reduce data conversion errors. Polynomial regression model verification is also used as a reference for future development of NIRS to automatically determine VT1 and VT2. This study investigated the reliability and validity of VT1 and VT2 determined via NIRS oxygenation and deoxygenation raw data using a polynomial regression model. This study’s hypotheses were that VT1 and VT2 determined using raw NIRS data and a polynomial regression model would have good reliability and viability.

Materials and methods

Subjects

This study was approved by the ethics committee of Chang Gung Memorial Hospital. The recruitment criteria for the male college students were (1) non-smokers, (2) no medication or vitamin use, and (3) no cardiopulmonary or hematological diseases. A total of 100 subjects were recruited and all provided informed consent. The subjects were instructed to refrain from exercise for ≥24 h before the exercise test. The subjects arrived at the testing center at 10:00 AM to eliminate any possible diurnal effect. All of the subjects completed the GXT once and 20 repeated the GXT after 48 h to confirm the test’s reliability.

Graded exercise test (GXT)

Each subject underwent a GXT to determine VT1 and VT2. The GXT was measured using a bicycle ergometer (Corvial 400, Lode). After 2 min of unloaded pedaling, the loading was increased by 30 W every 3 min until exhaustion (Wang et al., 2010, Mao et al., 2011). Heart rate (HR), blood pressure (BP), minute ventilation (VE), oxygen consumption (VO2), and CO2 production (VCO2) were collected using a biomedical system (Powerlab, USA). Arterial O2 saturation was monitored via finger pulse oximetry (model 9500, Nonin Onyx). The VO2max criteria were (1) the level of VO2 increased < 2 ml·kg−1·min−1 over ≥ 2 min, (2) HR exceeded the predicted 90% maximal heart rate (HRmax), and (3) the respiratory exchange ratio (RER) exceeded 1.2 (Wang et al., 2010, Mao et al., 2011).

Determination of VT1 and VT2

VT1 was determined by an increase in VE/VO2 with no increase in VE/VCO2 and departure from the linearity of ventilation (VE), whereas VT2 corresponded to an increase in both VE/VO2 and VE/VCO2 (Beaver et al., 1986). All of the VT1 and VT2 assessments were confirmed by visually inspecting graphs of the time plotted against each relevant respiratory variable. VT1 and VT2 were independently determined by two experienced exercise physiologists. When the results diverged, VT1 and VT2 were judged by a third exercise physiologist.

NIRS skeletal muscle hemodynamics

Local skeletal muscle hemodynamics profiles were simultaneously monitored using an NIRS system (Oxymon, Artinis, Netherlands). The transmitting and receiving optode was placed on the left vastus lateralis (VL) muscle (12 cm above the proximal border of the patella and 3–5 cm lateral to the midline of the thigh) (Wang et al., 2010). The distance between the optodes was adjusted to ensure proper placement (range: 2.5–3.5 cm) and good signal strength (10–30%) and the sampling frequency was set at 10 Hz. O2Hb and HHb were presumed to be reliable estimators of changes in tissue oxygenation or deoxygenation status representing regional O2 delivery and utilization. Total hemoglobin (tHb) reflected changes in the tissue blood volume (Boushel et al., 2001, DeLorey et al., 2003).

Determination of VT1NIRS and VT2NIRS

The following procedures were used to determine tissue NIRS reflecting VT1 and VT2: (1) 10 Hz O2Hb and HHb data were plotted from the beginning to the end of GXT; (2) the trend lines were drawn and adjusted using a polynomial regression model with Microsoft Excel; (3) VT1NIRS was determined by the peak of O2Hb polynomial function (the maximal value of O2Hb); and (4) VT2NIRS was determined by the peak of HHb polynomial function (the maximal value of HHb) (Fig. 1).

Fig. 1

NIRS signals changes during the GXT in one subject. Red dots: oxygenated hemoglobin. Blue dots: deoxygenated hemoglobin.

Statistical analysis

All of the cardiopulmonary data collected during the GXT were calculated at an average of 60 s. If there was an obvious deviation in the data, such as extreme values due to the subject sneezing or coughing, then that data were deleted. If the average exceeded 4 standard deviations, then the value was listed as an extreme value and deleted. The data were analyzed by SPSS for Windows 23.0. The independent t-test was used to compare VO2 at VT1NIRS and VT2NIRS between the first (T1) and second (T2) tests for reliability. The paired t-test was used to compare VO2 between VT1VET vs VT1NIRS and VT2VET vs VT2NIRS. Pearson’s correlation was used to evaluate the association of VO2 of VT1VET vs VT1NIRS and VT2VET vs VT2NIRS. A Bland-Altman analysis (Bland and Altman, 1986) was used to assess similarities between the ventilatory and NIRS methods. Statistical significance was α = 0.05.

Results

The subjects’ basic anthropometric and cardiopulmonary characteristics are presented in Table 1. To confirm the reliability of VT1NIRS and VT2NIRS during the GXT between T1 and T2, the correlations and similarities were assessed. There were significant positive correlations in VT1NIRS and VT2NIRS between T1 and T2. The Bland-Altman analysis demonstrated that the mean differences between VT1NIRS and VT2NIRS were all within a 95% (±1.96 SD) confidence interval (Fig. 2).

Table 1

The subjects’ basic anthropometric and cardiopulmonary characteristics.

Anthropometric
Age (years)	22 ± 3
Weight (kg)	67.1 ± 10.5
Height (cm)	173.1 ± 6.0
BMI (kg m−2)	22.3 ± 0.3
Cardiopulmonary	Baseline	Peak exercise
Workload (W)	0 ± 0	210 ± 5
HR (beats min−1)	77 ± 3	193 ± 2
VE (L min−1)	6.8 ± 0.5	127.5 ± 4.6
VO2 (mL kg−1 min−1)	3.7 ± 0.3	54.5 ± 1.3
RER	0.85 ± 0.01	1.18 ± 0.02
SpO2 (%)	96 ± 1	94 ± 1

Data are shown as mean ± SD. BMI: body mass index; HR: heart rate; VE: minute ventilation; VO2: oxygen uptake; RER: respiratory exchange rate; SPO2: saturation of peripheral oxygen.

Fig. 2

Reliability analysis between the first (T1) and second (T2) tests (n = 20).

There were significant positive correlations between the VO2 of VT1VET vs VT1NIRS and VT2VET vs VT2NIRS, validating VT1 and VT2 measured by the ventilatory and NIRS methods. The Bland-Altman analysis showed that the mean differences in VT1VET vs VT1NIRS and VT2VET vs VT2NIRS were consistent (Fig. 3). The comparisons of VO2 between VT1VET vs VT1NIRS and VT2VET vs VT2NIRS demonstrated that there were no significant differences between the ventilatory and NIRS methods (Table 2).

Fig. 3

Validity analysis of the ventilatory and NIRS methods (n = 100).

Table 2

Comparison of the cardiopulmonary parameters using the ventilatory and NIRS methods.

Parameters	VT1VET	VT1NIRS	VT2VET	VT2NIRS
Workload (W)	129 ± 3	122 ± 4	167 ± 3	159 ± 3
HR (beats min−1)	146 ± 1	144 ± 2	170 ± 1	165 ± 2
VE (L min−1)	60.9 ± 2.1	59.5 ± 2.2	93.0 ± 2.4	87.2 ± 2.2
VO2 (mL kg−1 min−1)	35.8 ± 1.0	35.0 ± 1.0	47.4 ± 1.0	46.2 ± 1.1
RER	1.01 ± 0.01	1.00 ± 0.01	1.09 ± 0.01	1.07 ± 0.01
SpO2 (%)	96.1 ± 0.1	96.1 ± 0.1	96.1 ± 0.1	95.9 ± 0.2

Data are shown as mean ± SD. *p < .05 calculated using the paired t-test to compare the ventilatory and NIRS methods. VT1VET: first ventilatory threshold determined by ventilatory methods; VT2VET: second ventilatory threshold determined by ventilatory methods; VT1NIRS: first ventilatory threshold determined by NIRS; VT2NIRS: second ventilatory threshold determined by NIRS; HR: heart rate; VO2: oxygen uptake; VE: minute ventilation; RER: respiratory exchange rate; SPO2: saturation of peripheral oxygen.

Discussion

This study’s main findings are: (1) skeletal muscle oxygenated and deoxygenated status detected via NIRS are reliable and valid for determining VT1 and VT2 using a mathematical polynomial regression model, (2) VT1NIRS reflected the limitation of muscle oxygenation during the GXT, and (3) VT2NIRS reflected the limitation of muscle deoxygenation during the GXT.

NIRS is a valid method that can be used to determine VT during exercise as previously confirmed in many populations, including normal subjects, children (Moalla et al., 2005), patients (Hamaoka et al., 2007), and athletes (van der Zwaard et al., 2016). NIRS was also confirmed as a non-invasive method of determining exercise intensity. Many studies demonstrated medium to high correlations between NIRS and traditional methods. The NIRS index demonstrated a decrease in oxyhemoglobin below baseline, a first sharp decrease in the oxygenation index, and first and second inflection early in the oxyhemoglobin slope (Bhambhani et al., 1997, Miura et al., 1998, Grassi et al., 1999, Terakado et al., 1999, Wang et al., 2006, Soller et al., 2008).

Most studies determined the NIRS index via intuitive or simple linear regression. Recent studies used more complex methods of determining the NIRS index, including tissue oxygen saturation using the Dmax method (Karatzanos et al., 2010), non-linear changes in cerebral oxygenation or deoxygenation (Racinais et al., 2014), and ΔO2HbMb-HHbMb by double linear regression (van der Zwaard et al., 2016). The present study used NIRS origin muscle O2Hb and HHb data to determine VT and RCP using a polynomial regression model. VTNIRS and RCPNIRS in this study were respectively 59.4% and 86.1%VO2max. Moalla et al. (2005) investigated VTNIRS using respiratory muscle oxygenation data during an incremental exercise cycling test. VTNIRS was 73.8 ± 6.9% and VTV-slope was 69.7 ± 6.8% of VO2max (Moalla et al., 2005). Karatzanos et al. (2010) studied gastrocnemius muscle oxygen saturation during an incremental treadmill test. VT was 86.2 ± 6.9% and NT (NIRS threshold) was 88.7 ± 7.3% VO2max (Karatzanos et al., 2010). NT was determined using a linear model. Racinais et al. (2014) detected VL muscle oxygenation during an incremental cycling test. Cerebral O2Hb was 56 ± 8% and HHb was 56 ± 13% at VT1. Cerebral O2Hb was 86 ± 8% and HHb was 87 ± 8% at VT2. Muscle oxygenation only related to VT2 of O2Hb was 78 ± 9% and HHb was 80 ± 8% (Racinais et al., 2014). van der Zwaard et al. (2016) used maximal incremental cycling to detect oxygenation breakpoints via double linear regression (the least combined residual sum of squares). ΔO2HbMb-HHbMb was 65 ± 11 %POpeak in male cyclists, 62 ± 12% %POpeak in female cyclists, 58 ± 12 %POpeak in endurance trained males, and 62 ± 10 %POpeak in recreationally trained males. ΔO2HbMb-HHbMb was not significantly different from VT1 but considerably lower than VT2 (van der Zwaard et al., 2016).

Our results were very similar to those using a non-linear regression model. This study also confirmed that VT1VET and VT1NIRS had a good correlation and VTNIR (59.4%) occurred earlier than VT (61.3%). As the exercise intensity increased, the oxygen deficit increased accompanied by anaerobic metabolism. Lactic acid in the blood began to accumulate and bicarbonate ions inside the muscle cells increased. Increasing intracellular carbon dioxide also led to an increase in CO2 in the blood, which was reflected by changing ventilation. The initiation of anaerobic metabolism detected by ventilation was defined as VT (Wasserman et al., 1981, Beaver et al., 1986). Our results showed that decreased muscle oxygenation (O2Hb) was a good predictor of VT1. Muscle O2Hb and HHb changed according to the oxyhemoglobin dissociation curve. Lactic acid, carbon dioxide, and hydrogen ions affected the oxyhemoglobin dissociation curve and increased the dissociation of oxygen from hemoglobin. A right shift of the Bohr effect indicated an decrease in O2Hb and a decrease in HHb in the muscle. Although VO2 of VT1VET and VT1NIRS almost coincided, VT1NIRS tended to occur earlier than VT1VET. Our findings confirmed that VT1NIRS determined using raw data with a polynomial regression is a good predictor of VT1VET.

This study also confirmed that VT2VET and VT2NIRS have good correlation. VT2NIRS (86.1%) occurred earlier than VT2VET (88.6%). The definition of VT2 was the starting point of respiratory compensation in response to exercise-induced metabolic acidosis. Beyond VT2, the increase in hydrogen ions enhanced a shift to the right of the oxyhemoglobin dissociation curve and further increased oxygen dissociation. In this study, VT2NIRS was determined by the peak of HHb function, which may have been caused by the dissociation of oxygen from myoglobin. VT2NIRS also formed due to attenuated O2Hb decrease. The decrease in O2Hb might have increased oxygen dissociation or decreased blood flow. Our non-published data indicated that cardiac output declined during the highest intensity exercise, which might explain why the muscle blood flow also decreased as shown in the present results.

Conclusions

In conclusion, NIRS muscle oxygenation and deoxygenation are reliable and valid methods of determining VT1 and VT2 using a mathematical polynomial function model. VT1NIRS reflected the limitation of muscle O2 delivery (O2Hb) and VT2NIRS reflected the limitation of muscle O2 utilization (HHb) during incremental cycling exercise.