Muscle oxygen saturation reflects individualized responses to exercise intensity during rowing ergometer exercise

Abstract

Background

Precise monitoring of training intensity in rowing requires indicators capturing both systemic and local physiological responses. Muscle oxygen saturation (SmO₂), measured by near-infrared spectroscopy (NIRS), reflects local muscular oxygen utilization and may complement traditional systemic indicators such as heart rate (HR), oxygen uptake (V̇O₂), and mechanical power output (P). This study aimed to investigate SmO₂ dynamics in upper- (biceps brachii, BB) and lower-limb (vastus lateralis, VL) muscles during incremental exercise, representative training sessions, and a 2000 m rowing ergometer race, and to evaluate its practical application for individualized monitoring.

Methods

Twenty-six trained male rowers (age: 19.96 ± 2.66 years; height: 189.12 ± 6.13 cm; body mass: 80.12 ± 9.21 kg) performed a graded incremental test, four representative training sessions, and a 2000 m race on a rowing ergometer. SmO₂ was measured continuously on biceps brachii and vastus lateralis using portable NIRS devices, along with HR, V̇O₂, and P. Linear mixed-effects models were used to examine within-subject associations between SmO₂ and systemic indicators, accounting for repeated measures across exercise stages.

Results

During incremental exercise, SmO₂-VL decreased from 73.7 ± 6.76% at rest to 7.88 ± 4.12% at maximal power, whereas SmO₂-BB decreased from 68.18 ± 4.49% to 28.93 ± 6.29%. Within-subject associations between SmO₂-VL and V̇O₂ (β = −1.255, 95% CI − 1.315 to − 1.194, p < 0.001), HR (β = −0.647, 95% CI − 0.675 to − 0.618, p < 0.001), and P (β = −0.186, 95% CI − 0.194 to − 0.177, p < 0.001) were significantly stronger than corresponding SmO₂-BB associations. SmO₂-VL also exhibited robust negative associations with HR (r = − 0.857, p < 0.001) and P (r = − 0.841, p < 0.001) during the 2000 m race. Variability in SmO₂ responses across training sessions reflected differences in exercise intensity.

Conclusion

SmO₂ decreases progressively with increasing exercise intensity, with the vastus lateralis showing greater sensitivity than the biceps brachii. Within-subject SmO₂-VL measurements provide a reliable indicator of local muscular oxygen extraction and can complement systemic variables for individualized training monitoring in rowing.

Introduction

Rowing is a strength-endurance sport, and athletic performance depends on aerobic and anaerobic capacity, rowing technique, and tactical execution [1, 2]. Aerobic capacity is particularly important in rowing [3], with aerobic energy contribution during a 2000 m race reported to reach approximately 86% in elite heavyweight male rowers [1, 4]. Aerobic performance depends on the integrated ability of the body to extract, transport, and utilize oxygen for energy production [5]. Therefore, monitoring oxygen-related physiological responses is essential for evaluating exercise intensity and guiding training prescription.

Commonly used indicators for monitoring rowing intensity include oxygen uptake (V̇O₂), heart rate (HR) and power output (P). These variables provide valuable information about whole-body physiological and mechanical load. However, they mainly reflect systemic responses and cannot directly indicate oxygen utilization within specific working muscles. This is important in rowing because the sport requires coordinated upper- and lower-limb actions, but the contribution of different muscle groups is not equal [6]. During rowing, the stroke consists of drive and recovery phases, with approximately 32 ~ 38 duty cycles per minute during a 2000 m race [1]. Previous biomechanical studies have shown that approximately 75% ~ 80% of rowing stroke power is generated by the legs [7, 8]. Therefore, monitoring local muscular responses in both the upper and lower limbs may provide additional information beyond systemic indicators.

Near-infrared spectroscopy (NIRS) provides a non-invasive method for assessing local muscle oxygenation. Early methods for measuring blood oxygen, such as gasometry and electrochemical techniques, were limited in their ability to assess oxygen dynamics within living tissues [9, 10]. With the development of spectroscopy and near-infrared light technology, NIRS made it possible to monitor tissue oxygenation in vivo. In 1977, Frans Jöbsis first proposed the use of NIRS to assess tissue oxygen parameters and the balance between oxygen supply and consumption [11]. In 1992, Britton Chance applied this technique to examine quadriceps muscle oxygen saturation recovery in rowers after training [12]. With the development of portable NIRS devices, muscle oxygen saturation (SmO₂) can now be measured continuously, non-invasively, and in real time during exercise, making it a promising tool for applied sports monitoring [13].

Previous studies have shown that SmO₂ is related to exercise intensity in several exercise modes. During incremental exercise, SmO₂ decreases progressively as intensity increases, and the point of rapid SmO₂ decline has been associated with the anaerobic threshold [14]. In highly trained cyclists, SmO₂ has shown significant associations with systemic intensity indicators such as V̇O₂, HR, and blood lactate [15]. In elite male cross-country skiers, SmO₂ reductions were intensity dependent during constant-load exercise [16]. During off-road time trials, SmO₂ has also been reported to provide useful information for monitoring running intensity and may serve as a more precise physiological marker for assessing running intensity during training and competition [17]. These findings suggest that SmO₂ may be a valuable local indicator of muscular oxygenation during exercise.

However, the rowing-specific application of SmO₂ remains insufficiently clarified. Existing studies have mainly examined SmO₂ in single exercise modes or isolated testing conditions, whereas fewer studies have connected incremental testing, training sessions, and race-like conditions within the same rowing population. In addition, although rowing involves both upper- and lower-limb muscle groups, it remains unclear whether SmO₂ measured at an upper-limb site and a lower-limb site provides equivalent information about rowing intensity. This issue is particularly relevant because the lower limbs contribute the majority of rowing power, whereas the upper limbs may play a more variable role in force transmission and stroke coordination.

Therefore, the biceps brachii (BB) and vastus lateralis (VL) were selected as representative upper- and lower-limb monitoring sites because of their accessibility for wearable NIRS assessment and their active involvement during rowing exercise [14, 18]. From a practical perspective, differences in SmO₂ responses between upper- and lower-limb muscles may provide information beyond that obtained from whole-body physiological indicators. Greater deoxygenation in a specific muscle group may indicate a higher local metabolic demand and a greater contribution to exercise performance. Consequently, identifying whether the biceps brachii or vastus lateralis shows a stronger relationship with exercise intensity may help coaches identify limiting muscle groups, optimize training emphasis, and better monitor localized fatigue and recovery during rowing training and competition.

Accordingly, this study aimed to investigate the relationships between SmO₂-BB, SmO₂-VL, V̇O₂, HR, and P during incremental rowing ergometer exercise, and to evaluate the practical utility of SmO₂-VL for monitoring exercise intensity during rowing training sessions and a 2000 m race. We hypothesized that SmO₂-VL would demonstrate stronger associations with systemic exercise intensity indicators than SmO₂-BB, reflecting the dominant contribution of the lower limbs to rowing power production. We further hypothesized that identifying these muscle-specific differences would provide practical information for monitoring localized training load in rowers.

Materials & methods

Participants

Twenty-six trained male rowers volunteered to participate in this study. Their basic characteristics are presented in Table 1. A priori power analysis was conducted using G*Power software (Version 3.1.9.7, University of Düsseldorf, Germany). Based on an expected correlation coefficient of r = 0.60, an α level of 0.05, and statistical power of 0.80, the minimum required sample size was 19 participants. Given that NIRS signals may be affected by actual subcutaneous adipose tissue thickness, participants with the actual subcutaneous adipose tissue thickness greater than 12 mm at either measurement site were excluded [19, 20]. To allow for possible attrition, 26 participants were finally recruited after screening [21].

Table 1.

Basic information of participants

N	Age (years)	Height (cm)	Weight (kg)	Training experience (years)	Vastus lateralis subcutaneous adipose tissue thickness (mm)	Biceps brachii subcutaneous adipose tissue thickness (mm)
26	19.96 ± 2.66	189.12 ± 6.13	80.12 ± 9.21	4.54 ± 2.08	4.88 ± 1.07	8.23 ± 2.23

Values are presented as mean ± SD

All participants were healthy, free from current injuries, and had no known contraindications to high-intensity exercise, including cardiovascular or skeletal muscle diseases. All participants were informed of the study procedures, potential risks, and requirements, and then written informed consent was obtained. The study was in accordance with the Declaration of Helsinki and approved by the Scientific Research Ethics Committee of Shanghai University of Sport (Approval No. 102772025RT148).

Methods

Test arrangement

This study used an observational repeated-measures design. The test arrangement is shown in Fig. 1. All participants completed three testing components over three consecutive weeks. All tests were performed on a rowing ergometer (Concept 2 RowErg, PM5 monitor, USA) under the same environmental conditions. The drag factor setting of the rowing ergometer remained unchanged throughout all tests.

Fig. 1.

Test arrangement

The incremental load test was conducted in the first week to determine the relationships among SmO₂-BB, SmO₂-VL, V̇O₂, HR, and P across increasing exercise intensity. In the second week, participants completed the selected training session tests to examine the practical application of SmO₂-VL during 4 representative rowing training conditions. In the third week, participants completed the 2000 m race test to assess SmO₂-VL responses under race-like high-intensity conditions. These tests were separated by approximately one week to allow sufficient recovery and to reduce the potential influence of fatigue from previous testing sessions.

An experimental briefing was held for all participants before the formal study, the purpose, schedule, procedures, and precautions of the study were explained in detail. Participants were then familiarized with the testing procedures to reduce potential learning effects during the formal tests. After that, anthropometric measurements were conducted. Body height and body mass were measured using a digital scale and stadiometer (Filizola, São Paulo, Brazil). Actual subcutaneous adipose tissue thickness was measured using a skinfold caliper (Innovare 2; Cescorf, Porto Alegre, Brazil) [22, 23]. Participants were instructed to avoid strenuous exercise before tests. Their diet and daily schedule were kept consistent under the unified management of the rowing team. They were also instructed to avoid smoking, caffeine intake, and any non-prescribed drugs before testing.

Incremental load test procedure

The order of the incremental load tests was randomly determined by drawing lots. Based on each participant’s previously assessed rowing ability, determined from their 2000 m rowing ergometer personal best, the initial workload was set at either 120–180 W. The test consisted of 2-min stages, with workload increased by 30 W at each stage without rest until exhaustion.

Exhaustion was determined using the following criteria [24–26]: (1) volitional exhaustion, defined as the inability to continue exercise or maintain the target power output despite strong verbal encouragement; (2) respiratory rate > 35 breaths/min; (3) maximum heart rate (HR_max) ≥ 95% of the age-predicted HR_max (220– age); and (4) respiratory exchange ratio > 1.1(15-s averaged value). The test was considered valid when at least three of these criteria were fulfilled simultaneously.

Before the test, participants were fitted with HR monitoring equipment, a gas metabolic analysis system, and NIRS sensors placed on the dominant-side biceps brachii and vastus lateralis. Participants then completed a standardized 15-min warm-up, including dynamic stretching, neuromuscular activation, and rowing ergometer exercise. During the warm-up, all devices were checked, and any incorrectly positioned sensors or equipment were adjusted immediately.

After a 5-min seated rest, the incremental load test began. HR, gas exchange variables, P, and SmO₂ signals were monitored continuously and non-invasively throughout the test. Strong verbal encouragement was provided by coaches and testers until participants reached exhaustion.

Training sessions tests and 2000 m race test procedure

Four representative rowing ergometer training sessions were selected in consultation with the athletes’ coaches: 1 set × 12,000 m, 2 sets × 6000 m, 1 set × 30 min, and 10 sets × 500 m. These sessions were chosen to represent different training demands and intensity levels commonly used in rowing training. The 2000 m race test was conducted according to standard rowing ergometer race procedures.

Consistent with the incremental load test, participants were fitted with HR monitoring equipment and a NIRS sensor placed on the dominant-side vastus lateralis. SmO₂-VL was selected for the training session and race tests because the incremental load test showed stronger associations with systemic/mechanical intensity indicators than SmO₂-BB. Participants then completed the same standardized 15-min warm-up used in the incremental load test. After a 5-min rest, the training session or 2000 m race test began.

During the training session tests, coaches regulated exercise intensity using real-time HR data in accordance with their usual training practice. SmO₂-VL and HR were recorded continuously and non-invasively throughout each session. During the 2000 m race test, maximal verbal encouragement was provided by coaches and testers to ensure maximal effort.

SmO₂ test method

SmO₂ was measured using a portable near-infrared spectroscopy device (Moxy Muscle Oxygen Monitor, Fortiori Design LLC, USA) [27]. The device measures 61 mm × 44 mm × 21 mm and weighs 48 g. It uses multiple wavelengths of light to estimate SmO₂ and provides data at a sampling frequency of 1 Hz. According to the device specifications, the measurement depth is approximately 12 mm; therefore, actual subcutaneous adipose tissue thickness at the measurement sites was assessed before testing to reduce the potential influence of subcutaneous adipose tissue thickness on NIRS signal quality.

To improve the standardization of wearable NIRS data acquisition, all sensor placements were performed by the same trained tester according to predefined anatomical landmarks. Before sensor placement, the skin was cleaned and dried. The biceps brachii site was located near the midpoint of the contracted muscle belly, and the vastus lateralis site was located approximately 10 ~ 12 cm above the patellar tuberosity [28]. The sensor was positioned over the muscle belly, aligned with the long axis of the muscle, and secured with black self-adhesive bandage to minimize movement artefacts and external light interference [29]. After fixation, participants remained still until SmO₂ values stabilized. Signal quality was visually inspected before each test, and any unstable signal, visible sensor displacement, or abnormal SmO₂ trace was corrected before formal data collection. During rowing exercise, the sensor fixation was checked whenever abnormal signal fluctuations were observed.

HR test method

HR was recorded using a chest-strap monitoring system (Firstbeat Technologies Ltd., Jyväskylä, Finland). Before placement, the contact area of the chest strap was cleaned with alcohol-soaked cotton swabs to improve signal quality. The strap was positioned around the lower chest, with the sensor placed near the lower sternum/xiphoid process region. Strap tightness was adjusted according to each participant’s body shape and comfort to prevent movement or detachment during rowing exercise. Before each test began, the HR signal was checked to ensure stable data transmission. Formal testing started only after HR values were stable.

V̇O₂ test method

V̇O₂ was only measured during the incremental load test using a portable cardiorespiratory testing system (Cortex MetaMax 3B-R2, Leipzig, Germany). The system consisted of a breathing mask and gas analyzer, and gas exchange variables were recorded continuously using breath-by-breath measurement. Before each test, the system was warmed up and calibrated according to the manufacturer’s instructions [30]. The breathing mask was selected according to each participant’s face size to ensure an appropriate fit. Before placement, the mask and tubing were connected and checked for air leakage by blocking the outlet and confirming that the system remained sealed. After the device was fitted, participants performed several low-intensity movements to confirm the stability of the equipment and signal quality.

During the incremental load test, V̇O₂, respiratory exchange ratio, minute ventilation, respiratory rate, and other gas exchange variables were monitored continuously. If abnormal physiological responses or equipment-related problems occurred, the test was terminated immediately.

Statistical analysis

During the incremental load test, G0 represented the resting condition, G1–G10 represented successive exercise stages, and G11 represented the maximal or minimal value achieved for each variable by each participant. A stage was included in the analysis only when more than half of the stage duration was completed. For each stage, the average value of the final 30 s was used as the representative value for SmO₂, HR, and V̇O₂. The highest completed workload was defined as P_max. V̇O_2peak was determined as the highest 15-s averaged V̇O₂ value, and HR_max was defined as the highest HR value recorded during the test.

Stage values were expressed relative to individual maximal values as %V̇O_2peak, %HR_max, and %P_max. Based on %V̇O_2peak, five relative intensity zones were defined (50–59%, 60–69%, 70–79%, 80–89%, and 90–99% V̇O_2peak). Corresponding group-level mean values of %HR_max, %P_max, SmO₂-BB, and SmO₂-VL were calculated for each zone [31, 32].

For the four training sessions and the 2000 m race, 1 set × 12,000 m and 2 sets × 6000 m were each divided into 12 × 1000 m stages, the 1 set × 30 min session into 5 × 6 min stages, the 10 sets × 500 m session into 10 × 500 m stages, and the 2000 m race into 4 × 500 m stages. The average value of the final 30 s of each stage was used as the representative value for SmO₂-VL and HR.

All statistical analyses were performed using SPSS version 25.0 (IBM Corp., Armonk, NY, USA). Data normality was assessed using the Shapiro–Wilk test. Data are presented as mean ± SD. Statistical significance was set at p < 0.05.

Because SmO₂-BB and SmO₂-VL represent physiologically distinct muscle sites, analyses focused on site-specific oxygenation responses and their relationships with systemic and mechanical indicators rather than agreement or interchangeability between muscles.

To account for repeated observations across exercise stages, linear mixed-effects models (LMMs) were used to examine the associations between SmO₂-BB/SmO₂-VL and V̇O₂, HR, and P during the incremental load test. Participant was included as a random effect. To distinguish intra-individual responses from inter-individual variability, predictors were decomposed into within-subject and between-subject components using participant-specific means and deviations from those means [33]. Separate models were fitted for V̇O₂, HR, and P.

Because the initial workload differed between participants, stage number was not treated as a common absolute intensity level. Instead, analyses were based on the actual physiological and mechanical values measured at each stage.

For training sessions and the 2000 m race, Pearson correlation coefficients were calculated within each exercise condition to describe session-specific associations between SmO₂-VL and HR/P. Differences in correlation strength across exercise conditions were evaluated using Fisher’s Z transformation and chi-square homogeneity tests.

A separate LMM was used to examine the effects of training type and exercise stage on SmO₂-VL during training sessions. Training type, stage, and their interaction were included as fixed effects, while participant was included as a random effect. Stage was treated as a categorical variable because stage duration and physiological meaning differed among training protocols.

Results

All participants completed the study, and no one dropped out. Figure 2 presents representative data from one participant during the incremental load test. As workload increased progressively, both SmO₂-BB and SmO₂-VL showed a gradual decline, indicating increased local muscle deoxygenation with increasing exercise intensity (Fig. 2A). In contrast, HR and V̇O₂ increased progressively across stages, reflecting the expected elevation in systemic cardiovascular and metabolic demand (Fig. 2B). These representative responses illustrate the opposite trends between local muscle oxygen saturation and systemic intensity indicators during incremental rowing exercise.

Fig. 2.

Change of SmO₂-BB, SmO₂-VL and P (A), HR and V̇O₂ (B) during the incremental load test (taking one participant as an example)

Table 2 presents the physiological and mechanical responses across the incremental load test. As exercise intensity increased from G0 to G10, P, HR, and V̇O₂ increased progressively, whereas SmO₂-BB and SmO₂-VL decreased markedly. SmO₂-BB declined from 68.18 ± 4.49% at rest (G0) to 28.93 ± 6.29% at G10, representing a reduction of approximately 57.6%. Similarly, SmO₂-VL decreased from 73.7 ± 6.76% to 7.88 ± 4.12%, corresponding to an 89.3% reduction. In contrast, V̇O₂ increased from 11.51 ± 2.07 to 72.81 ± 5.15 mL/kg/min, HR increased from 86.14 ± 10.81 to 190.96 ± 3.6 beats/min, and P increased from 0 to 390 W. Compared with SmO₂-BB, SmO₂-VL exhibited a steeper decline throughout the incremental protocol, particularly at higher exercise intensities (G5–G10). The larger reduction observed in SmO₂-VL compared with SmO₂-BB suggests distinct site-specific oxygenation responses between the lower- and upper-limb muscles during incremental rowing exercise.

Table 2.

Physiological and mechanical responses across incremental load test stages

Phase	SmO2-BB (%)	SmO2-VL (%)	V̇O2 (mL/kg/min)	HR (beats/min)	P (watt)
G0	68.18 ± 4.49	73.7 ± 6.76	11.51 ± 2.07	86.14 ± 10.81	0 ± 0
G1	47.1 ± 8.33	52.26 ± 10.73	36.57 ± 4.29	130.19 ± 12	156.92 ± 29.77
G2	44.13 ± 8.75	46.69 ± 11.03	41.44 ± 3.91	140.44 ± 12.63	186.92 ± 29.77
G3	40.06 ± 9.65	38.87 ± 12.46	46.01 ± 3.67	150.88 ± 12.08	216.92 ± 29.77
G4	36.69 ± 9.81	29.36 ± 12.12	50.41 ± 3.95	160.46 ± 11.85	246.92 ± 29.77
G5	33.26 ± 8.7	22.37 ± 11.21	54.31 ± 3.62	168.45 ± 10.57	276.92 ± 29.77
G6	30.51 ± 8.33	16.57 ± 9.07	58.83 ± 5.05	176.2 ± 9.2	306.92 ± 29.77
G7	28.83 ± 7.89	12.01 ± 7.6	61.82 ± 5	182.28 ± 7.58	336 ± 30
G8	29.01 ± 6.35	10.74 ± 6.58	64.07 ± 5.35	185.49 ± 6.03	356.67 ± 30.68
G9	28.73 ± 6.96	9.44 ± 5.69	66.71 ± 5.86	186.31 ± 3.14	380 ± 30
G10	28.93 ± 6.29	7.88 ± 4.12	72.81 ± 5.15	190.96 ± 3.6	390 ± 0
G11	25.84 ± 6.32	9.89 ± 6.25	65.31 ± 6.32	189.77 ± 5.46	372.69 ± 27.06

Values are presented as mean ± SD

Linear mixed-effects models were used to examine the within-subject and between-subject associations between SmO₂ and physiological/mechanical variables during the incremental load test (Table 3). For SmO₂-VL, significant negative within-subject associations were observed with HR (β = −0.647, 95% CI: −0.675 to − 0.618, p < 0.001), V̇O₂ (β = −1.255, 95% CI: −1.315 to − 1.194, p < 0.001), and P (β = −0.186, 95% CI: −0.194 to − 0.177, p < 0.001). However, no significant between-subject associations were detected for HR (p = 0.987), V̇O₂ (p = 0.805), or P (p = 0.692). Similarly, SmO₂-BB showed significant negative within-subject associations with HR (β = −0.113, 95% CI: −0.118 to − 0.107, p < 0.001), V̇O₂ (β = −0.766, 95% CI: −0.801 to − 0.730, p < 0.001), and P (β = −0.113, 95% CI: −0.118 to − 0.107, p < 0.001). No significant between-subject associations were observed for any predictor (all p > 0.05).

Table 3.

Within- and between-subject associations between SmO2 and physiological and mechanical variables during the incremental load test

Outcome	Predictor	Effect	β	SE	df	t	95% CI	p
SmO2-BB	V̇O2	Within	-0.766	0.018	210.958	-42.520	-0.801, -0.730	< 0.001**
		Between	-0.009	0.332	23.962	-0.028	-0.696, 0.677	0.978
SmO2-BB	HR	Within	-0.113	0.003	210.953	-43.247	-0.118, -0.107	< 0.001**
		Between	-0.076	0.062	23.882	-1.229	-0.203, 0.051	0.231
SmO2-BB	P	Within	-0.113	0.003	210.953	-43.247	-0.118, -0.107	< 0.001**
		Between	-0.076	0.061	23.882	-1.229	-0.203, 0.051	0.231
SmO2-VL	V̇O2	Within	-1.255	0.031	210.913	-41.074	-1.315, -1.194	< 0.001**
		Between	-0.100	0.402	23.916	-0.250	-0.930, 0.729	0.805
SmO2-VL	HR	Within	-0.647	0.015	210.921	-44.540	-0.675, -0.618	< 0.001**
		Between	-0.004	0.230	24.003	-0.016	-0. 479, 0.472	0.987
SmO2-VL	P	Within	-0.186	0.004	210.921	-44.057	-0.194, -0.177	< 0.001**
		Between	0.031	0.077	23.801	0.400	-0.127, 0.189	0.692

^**p < 0.001

Overall, these findings indicate that changes in SmO₂ were primarily driven by within-athlete increases in exercise intensity during the incremental load test, whereas differences between athletes contributed little to the observed relationships. Moreover, the larger within-subject coefficients observed for SmO₂-VL compared with SmO₂-BB suggest that vastus lateralis oxygenation was more responsive to changes in exercise intensity than biceps brachii oxygenation.

Exploratory group-level SmO₂ reference values across %V̇O_2peak intensity zones are presented in Table 4. SmO₂-BB and SmO₂-VL decreased progressively as %V̇O_2peak increased. However, the standard deviations, particularly for SmO₂-VL, were relatively large across several zones, indicating substantial inter-individual variability in local muscle oxygenation responses.

Table 4.

Exploratory group-level SmO₂ reference values across %V̇O_2peak intensity zones during the incremental load test

Intensity Zones	%V̇O2peak (%)	SmO2-BB (%)	SmO2-VL (%)	%HRmax (%)	%Pmax (%)
50 ~ 59 (%)	54.64 ± 2.99	47.52 ± 7.65	49.70 ± 6.85	69.06 ± 5.12	42.25 ± 6.40
60 ~ 69 (%)	64.63 ± 2.90	44.79 ± 6.65	44.53 ± 13.09	75.12 ± 5.15	51.57 ± 6.11
70 ~ 79 (%)	74.64 ± 2.59	37.82 ± 8.28	34.02 ± 14.86	81.34 ± 5.19	62.02 ± 6.68
80 ~ 89 (%)	85.26 ± 3.02	32.57 ± 8.88	21.28 ± 12.76	90.01 ± 3.84	75.91 ± 7.37
90 ~ 99 (%)	96.01 ± 2.71	28.27 ± 6.98	12.60 ± 8.20	97.11 ± 2.66	92.64 ± 6.86

Values are presented as mean ± SD

As shown in Table 5, SmO₂-VL exhibited significant negative correlations with HR and P across all training sessions and the 2000 m race (p < 0.001). During the four training sessions, correlation coefficients ranged from low to high (r = − 0.379 to − 0.810), whereas consistently high negative correlations were observed during the 2000 m race (HR: r = − 0.857; P: r = − 0.841). Fisher’s Z-transformed correlation coefficients indicated variability in the strength of associations across exercise conditions. The homogeneity test revealed that correlations between SmO₂-VL and HR differed significantly across sessions (χ² = 12.77, p = 0.012), while correlations between SmO₂-VL and P showed a non-significant trend toward heterogeneity (χ² = 8.81, p = 0.066). The linear mixed-effects model demonstrated significant main effects of training type (F = 10.313, p < 0.001) and stage (F = 76.104, p < 0.001) on SmO₂-VL. However, the interaction effect between type and stage was not significant (F = 0.833, p = 0.711), indicating that although SmO₂-VL differed across training sessions and stages, the pattern of change across stages was similar among the different training types.

Table 5.

Correlations between SmO₂-VL, HR and P across training sessions and the 2000 m race

Training Sessions/Race	SmO2-VL × HR			SmO2-VL × P
	r	Z		r	Z
1 set × 12,000 m	-0.379**		-0.399	-0.453**		-0.488
2 sets × 6000 m	-0.519**		-0.575	-0.635**		-0.750
1 set × 30 min	-0.787**		-1.064	-0.802**		-1.104
10 sets × 500 m	-0.789**		-1.069	-0.810**		-1.127
2000 m race	-0.857**		-1.282	-0.841**		-1.225

^**: p < 0.001. Correlation coefficients were transformed using Fisher’s Z transformation

Discussion

The present study investigated the behavior of SmO₂ during incremental rowing exercise, representative training sessions, and a 2000 m rowing ergometer race. The main findings were: (1) both SmO₂-BB and SmO₂-VL decreased progressively with increasing exercise intensity, while V̇O₂, HR, and P increased; (2) significant negative within-subject associations were observed between SmO₂ and V̇O₂, HR, and P, whereas between-subject associations were negligible; (3) SmO₂-VL exhibited stronger responses to exercise intensity than SmO₂-BB; and (4) SmO₂-VL remained significantly associated with HR and P across training sessions and competition, supporting its potential utility as an individualized indicator of local physiological stress.

The progressive decline in SmO₂ observed during the incremental load test is consistent with the physiological role of SmO₂ as an indicator of the balance between local oxygen delivery and oxygen utilization, which is consistent with numerous previous studies [15, 28, 34, 35]. As exercise intensity increased, V̇O₂, HR, and P rose continuously, reflecting increasing systemic cardiovascular and metabolic demands, whereas SmO₂ decreased markedly, indicating progressively greater oxygen extraction within the working musculature [36].

During the incremental protocol, SmO₂-BB decreased by approximately 57.6%, whereas SmO₂-VL decreased by approximately 89.3%, demonstrating substantial local deoxygenation as exercise intensity approached maximal levels. These findings are consistent with previous studies showing that NIRS-derived SmO₂ is sensitive to changes in exercise intensity and provides information distinct from systemic variables such as HR and V̇O₂ [32, 37]. While HR and V̇O₂ primarily reflect whole-body physiological responses, SmO₂ directly reflects local muscular oxygen extraction and therefore provides insight into the physiological state of specific working muscles.

This distinction is important for interpreting SmO₂ in rowing. As oxygen is transported through the respiratory and circulatory systems and ultimately used in skeletal muscle mitochondria, local SmO₂ reflects the final peripheral stage of oxygen delivery and utilization. Changes in SmO₂ therefore reveal local oxygen dynamics that may not be fully captured by systemic indicators alone. In this sense, the value of SmO₂ monitoring lies not only in its relationship with established intensity markers, but also in its ability to provide additional information about regional muscular responses during rowing-specific exercise.

A major finding of the present study was that SmO₂-VL demonstrated consistently stronger associations with exercise intensity indicators than SmO₂-BB. During the incremental test, the within-subject association between SmO₂-VL and V̇O₂ (β = −1.255) was substantially stronger than that observed for SmO₂-BB (β = −0.766). Similar differences were observed for HR and P. It suggests that the underlying physiological mechanisms governing oxygen utilization may differ between upper- and lower-limb muscles [38]. The vastus lateralis, as a primary locomotor muscle, undergoes greater oxygen extraction during incremental exercise, whereas the biceps brachii exhibits a comparatively attenuated response. These findings indicate that SmO₂ should be interpreted as a local rather than global physiological marker. Measurements obtained from different muscle sites are not interchangeable and should be selected based on the specific demands of the exercise task. It should be noted that biceps brachii and vastus lateralis were selected as representative monitoring sites rather than comprehensive indicators of upper- and lower-limb oxygenation. Rowing performance involves coordinated contributions from multiple muscle groups, including the latissimus dorsi, gluteus maximus, and hamstring muscles, which were not monitored in the present study. From an applied perspective, these results highlight the importance of site selection in wearable muscle oxygenation monitoring and suggest that multi-site assessment may provide more comprehensive insights into exercise physiology.

Previous studies have also reported heterogeneous SmO₂ responses across muscle groups. In a ramp cycling exercise study, SmO₂ was measured in several lower-limb muscles, including the vastus lateralis, rectus femoris, vastus medialis, biceps femoris, gastrocnemius lateralis, gastrocnemius medialis, and tibialis anterior. That study showed that deoxygenation increased with exercise intensity, but the magnitude and pattern of SmO₂ responses differed among muscles [39]. Although cycling involves less upper-limb contribution than rowing, the finding supports the broader principle that SmO₂ responses are muscle specific rather than uniform across the body. More directly relevant evidence comes from simulated double poling in elite male cross-country skiers. In that study, SmO₂ responses of the triceps brachii and vastus lateralis were examined during constant-load exercise at different intensities. The authors found differences between upper- and lower-limb muscle oxygenation responses, as well as distinct desaturation patterns during maximal exercise [16]. During rowing, the upper and lower limbs work in a coordinated sequence, and the biceps brachii of the upper limb and vastus lateralis of the lower limb are both active during the rowing stroke [40–42]. However, their functional contributions are not equivalent. During the drive phase, leg extension provides the major source of propulsion, and the force and contraction velocity of the quadriceps muscle group are critical for power production [3].

This study supports the general biomechanical principle that upper-body mechanical output depends on segmental mechanics. The role of the biceps brachii during rowing should not be interpreted as an isolated source of mechanical power. Rather, the upper limb contributes to force transmission from the handle to the trunk and assists in coordinating the pulling phase of the stroke. From a biomechanical perspective, upper-body mechanical output depends on segmental coordination, joint motion, and force transmission across linked body segments. Rigid-body mechanical models of explosive upper-body actions further indicate that upper-body power output is influenced by movement geometry and segmental mechanics rather than by a single muscle alone [43]. The variability of SmO₂-BB may also be influenced by stroke frequency and rowing strategy. In cyclic water sports, stroke frequency can alter coordination patterns, force transmission, and localized fatigue development [44]. During rowing, a given P may be achieved through different force–velocity strategies. For example, a higher stroke rate with lower force per stroke may impose different upper-limb coordination and metabolic demands compared with a lower stroke rate with greater force per stroke. Therefore, SmO₂-BB variability may partly reflect differences in stroke frequency, pulling strategy, and upper-limb neuromuscular coordination. However, because stroke frequency was not directly included in the present analysis, this interpretation should be considered a possible explanation rather than a confirmed mechanism. In addition, SmO₂-BB should not be interpreted as a comprehensive indicator of upper-body pulling capacity. Upper-limb pulling performance involves coordinated activation of multiple muscles and may vary substantially across individuals. Evidence from arm-dominant functional pulling tasks, such as rope climbing, also indicates that upper-body pulling performance and reliability can be variable [45].

From a physiological perspective, the stronger association between SmO₂-VL and whole-body intensity may reflect the greater metabolic involvement of the lower limbs during rowing. As exercise intensity increases, the quadriceps are required to generate repeated high-force contractions during the drive phase. This increases local oxygen extraction in the vastus lateralis and makes SmO₂-VL more closely linked to mechanical power output and systemic oxygen demand. In contrast, the biceps brachii contributes to force transmission and upper-limb pulling, but its relative contribution to total power production is smaller and more dependent on rowing technique [46]. Therefore, SmO₂-BB may be more affected by inter-individual differences in stroke strategy, arm involvement, and coordination, and the weaker association between SmO₂-BB and systemic/mechanical intensity indicators may partly reflect the fact that biceps brachii oxygenation captures only one local component of a broader and more variable upper-body pulling system.

One of the most important findings of this study emerged from the mixed-effects analyses. Significant associations between SmO₂ and exercise intensity indicators were observed almost exclusively at the within-subject level, whereas no significant between-subject associations were detected. This distinction has important practical implications. The absence of between-subject effects suggests that absolute SmO₂ values should be interpreted cautiously when comparing different athletes. Physiological factors such as muscle fiber composition, capillary density, local oxidative capacity, training status of the individual and sensor placement may contribute to substantial inter-individual variability in SmO₂ values [47]. Consequently, a given SmO₂ value may have different physiological meanings across athletes. In contrast, the strong within-subject relationships indicate that changes in SmO₂ within the same athlete provide meaningful information regarding changes in exercise intensity and local muscular stress. These findings support the use of SmO₂ as an individualized monitoring variable rather than as a universal metric for direct comparison between athletes.

Current ACSM guidelines classify exercise intensity using systemic variables such as %HR_max and %V̇O_2max [48]. These markers remain useful for general cardiorespiratory intensity prescription. While training zones based on V̇O₂, HR, and P are well established, they do not provide information about local muscular oxygenation. Our findings suggest that SmO₂ may offer additional value by revealing intra-session variability and local physiological responses that are not reflected in traditional intensity markers. The purpose of establishing SmO₂-BB and SmO₂-VL reference ranges in the present study is therefore not to replace existing HR-, V̇O₂-, or P- based training zones, but to supplement them with local muscle-specific information during rowing ergometer exercise. In this sense, SmO₂-based ranges may help coaches and athletes identify whether a given systemic training zone is accompanied by different levels of local muscular oxygen extraction.

The relatively large standard deviations observed in the SmO₂ values across %V̇O_2peak zones indicate marked inter-individual variability in local muscle oxygenation. This variability limits the use of group-based SmO₂ ranges as fixed training prescription zones. Therefore, the SmO₂ values reported in this study should be interpreted as exploratory reference values rather than universal thresholds. In practice, SmO₂ monitoring may be more useful when applied longitudinally within the same athlete, using individualized baseline values and athlete-specific response profiles.

Since direct V̇O₂ measurement is difficult to implement outside the laboratory, HR continues to be a convenient and widely used field-based monitoring tool. However, HR mainly reflects systemic cardiovascular strain and cannot directly indicate how oxygen delivery and utilization are distributed within specific working muscles. SmO₂-VL therefore provides additional information about peripheral muscular demand, particularly in the lower limb musculature that contributes substantially to rowing power production.

Another key finding of this study is that the strength of association between SmO₂-VL and HR, P differed significantly across training sessions and race. The homogeneity test confirmed that correlations were not consistent. This suggests that the coupling between local muscle oxygenation and systemic responses is modulated by exercise modality and intensity. In the training sessions and 2000 m race, significant negative correlations were observed between SmO₂-VL, HR and P, but the strength of these correlations varied across exercise conditions. The correlations were weaker during lower-intensity continuous training (e.g., 1 set×12000 m) and stronger during higher-intensity sessions (10 sets×500 m) and the 2000 m race. This pattern suggests that the coupling between local vastus lateralis oxygenation and systemic or mechanical intensity becomes stronger as exercise intensity increases. At lower intensities, oxygen delivery, muscle recruitment, and whole-body coordination may remain relatively stable, producing weaker associations between SmO₂-VL, HR and P. At higher intensities, increased quadriceps recruitment and greater local oxygen extraction may make SmO₂-VL more closely aligned with whole-body cardiovascular and mechanical demand.

This interpretation is supported by the homogeneity analysis of correlations across sessions. Fisher’s Z-based homogeneity testing showed that the correlations between SmO₂-VL and HR differed significantly across sessions, while the correlations between SmO₂-VL and P showed a non-significant trend. These results indicate that the relationship between local muscle oxygenation and systemic/mechanical intensity is not fixed across training contexts. These differences can be interpreted from a physiological perspective. During low-intensity, steady-state exercise, muscle oxygenation is regulated by multiple factors including local perfusion, metabolic efficiency, and fiber recruitment patterns, which may weaken its association with global cardiovascular indicators [49–51]. In contrast, during high-intensity or interval-based sessions, the increasing metabolic demand leads to tighter integration between local oxygen extraction and systemic cardiovascular responses, resulting in stronger correlations. SmO₂-VL should be interpreted as a context-dependent marker whose practical value may be greater during high-intensity rowing sessions, such as interval training and race-like efforts. Although stronger associations between SmO₂-VL and HR/P were observed during higher-intensity sessions, this should not be interpreted as indicating unlimited sensitivity of SmO₂-VL at maximal exercise intensities. As local oxygen extraction approaches its physiological maximum, SmO₂ may reach a lower plateau or floor, limiting its ability to further discriminate changes in intensity near V̇O_2peak. Therefore, SmO₂-VL may be most informative for identifying the transition toward high local muscular oxygen extraction and monitoring the development of local fatigue, whereas its interpretation at near-maximal intensity should be combined with HR, P.

The variability in correlations across training sessions should also be interpreted in relation to rowing mechanics. Different training formats may require distinct stroke rates, force–velocity strategies, and neuromuscular recruitment patterns. For example, long-distance continuous sessions such as 1 set × 12,000 m may be performed with relatively stable pacing and lower stroke frequency, whereas 10 sets × 500 m and 2000 m race efforts may involve higher stroke frequency, greater force application, and greater recruitment of the lower-limb musculature. Therefore, the stronger coupling between SmO₂-VL and HR/P during higher-intensity sessions may reflect both increased metabolic demand and altered rowing mechanics. Since stroke rate and drag factor were not included in the present analysis, these interpretations should be regarded as plausible mechanisms rather than directly tested effects.

Practical implications

The present findings suggest that SmO₂-VL may provide useful information for training monitoring in rowing. Unlike HR, which reflects cardiovascular stress, or P, which represents external workload, SmO₂ provides direct information regarding local muscle oxygen utilization. Therefore, monitoring SmO₂-VL may help coaches identify situations in which local muscular stress is disproportionately high relative to systemic responses.

However, SmO₂ should not be considered a replacement for HR or power monitoring. Instead, its greatest value likely lies in complementing traditional monitoring tools. The combination of external load (P), systemic load (HR), and local muscular load (SmO₂) may provide a more comprehensive understanding of an athlete’s physiological response to training and competition.

The exploratory SmO₂ reference values reported in this study may serve as preliminary benchmarks for future research. Nevertheless, the relatively large standard deviations observed across several intensity zones indicate substantial inter-individual variability, particularly for SmO₂-VL. Accordingly, these values should be interpreted as group-level reference ranges rather than universal training zones for prescription purposes.

Limitations

Several limitations should be acknowledged. First, although participants were screened using a skinfold thickness criterion consistent with the optical penetration depth of the NIRS device, NIRS measurements remain limited to superficial muscle tissues and may not fully represent oxygenation dynamics in deeper muscle layers. Therefore, the reported SmO₂ values should be interpreted as indicators of localized superficial muscle oxygenation.

Second, substantial inter-individual variability was observed, particularly within the V̇O_2peak-derived intensity zones. This variability likely reflects differences in muscle morphology, oxidative capacity, and rowing technique and limits the direct application of population-based SmO₂ thresholds.

Third, biomechanical variables such as stroke rate, drag factor, stroke length, and force–velocity characteristics were not recorded. Because similar P can be achieved using different rowing strategies, these factors may have influenced local oxygenation responses and should be incorporated into future investigations.

Fourth, the present study focused on the biceps brachii and vastus lateralis as representative upper- and lower-limb muscles. Other muscles that contribute substantially to rowing performance, including the latissimus dorsi and gluteus maximus, were not assessed and may provide additional insight into rowing-specific oxygenation responses.

Fifth, SmO₂ responses may approach a physiological floor at near-maximal exercise intensities when local oxygen extraction becomes extremely high. Consequently, the sensitivity of SmO₂ to further increases in intensity may be reduced near V̇O_2peak.

Sixth, this study focused exclusively on trained male rowers. Future research should expand the scope to include female rowers and athletes of varying competitive levels to evaluate sex-related differences and experience-related changes in SmO₂ responses.

Finally, this study indicates that SmO₂ is more appropriate for individualized longitudinal monitoring than for direct comparison of absolute values. Future studies should further examine the change in muscle oxygenation (ΔSmO₂) from individualized resting baselines to improve comparability.

Conclusion

SmO₂ measured by wearable NIRS devices decreases progressively as exercise intensity increases during rowing exercise and demonstrates strong within-subject associations with V̇O₂, HR, and P. SmO₂-VL appears more responsive to exercise intensity than SmO₂-BB, likely reflecting the greater contribution of the lower limbs to rowing power production. These findings suggest that SmO₂-VL may serve as a useful individualized indicator of local muscular stress during rowing training and competition when used alongside traditional measures such as HR and P.

Authors' contributions

J.Y. and B.G. carried out the conception and design of this study; J.Y. and Z.M. were responsible for the research and data analysis; J.Y. wrote the manuscript; Z.M. and B.G. supervised the study and revised the manuscript. All authors have read and agreed to the published version of the manuscript.

Funding

This work was supported by the Shanghai University of Sports under the Shanghai Key Laboratory of Human Movement Development and Protection funded project (11DZ2261100).

Data availability

The datasets used and/or analyzed during the current study are available from the first and corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

The study was approved by the Scientific Research Ethics Committee of Shanghai University of Sport (Approval No. 102772025RT148). Informed consent was obtained from all participants, who voluntarily took part in this experiment.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.