Functional Exercise Induces Adaptations in Muscle Oxygen Saturation in Division One Collegiate Butterfly Swimmers: A Randomized Controlled Trial

Abstract

This study investigates the impact of a five-week functional exercise intervention designed to enhance the muscular endurance of the posterior shoulder musculature, aiming to mitigate shoulder fatigue and overuse injury. Twelve Division I collegiate butterfly swimmers were recruited and evenly randomized into exercise (EX) and control (CTRL) groups. Weekly 100-yard butterfly sprints were performed, with Muscle Oxygen Saturation (SmO₂) continuously monitored using a wearable near-infrared spectroscopy (NIRS) device. This study is among the first to utilize wearable NIRS devices to monitor SmO₂ underwater during swimming, demonstrating that a targeted 5-week exercise program significantly improves posterior shoulder endurance, as evidenced by increased Posterior Shoulder Endurance Test (PSET) scores and distinctive SmO₂ adaptations in the EX-group compared to the CTRL group. These findings suggest that targeted dryland exercises can enhance posterior shoulder endurance with long-term implications for potentially reducing injury risk and improving performance.

1. Introduction

Swimming is a sport in which the athlete’s technique and power must work together seamlessly to propel themselves through the water while maximizing hydrodynamics [1]. Fatigue in swimming causes a chain reaction of inefficient compensatory muscle activation to occur, often resulting in form breakdown and less efficient movement patterns [2]. One apparent alteration in stroke mechanics occurs during the recovery phase of the butterfly stroke. This phase is traditionally characterized as the portion of the stroke in which the swimmer’s arms sweep above the surface of the water to initiate the subsequent stroke cycle. When a butterfly swimmer fatigues, there is a drastic reduction in hand velocity above the water, which can be observed across all levels of competition [3,4]. Studies using electromyography (EMG) have shown that the deltoid, rotator cuff, and rhomboid muscles are primarily responsible for the movement of the arms during the recovery phase of the butterfly stroke [5,6]. Shoulder injuries account for between 17% and 76% of all swimming-related injuries, with swimmer’s shoulder coined to represent a wide range of shoulder pathologies, many of which are initiated by the fatigue of the shoulder stabilizers [7,8]. This occurs when the fatigued rotator cuff muscles—such as the infraspinatus, supraspinatus, subscapularis, and teres minor—are unable to stabilize the glenohumeral joint, altering both scapular and glenohumeral kinematics [9]. Given that the butterfly style displays significantly higher activation and fatigue of these stabilizers compared to other strokes, there exists a critical gap in current training protocols that do not effectively address these weaknesses [6,10,11,12].

The implementation of wearable devices in swimming has only recently begun, with device engineering facing significant challenges due to the aquatic nature of the sport. The attenuation of wireless signals in water complicates real-time data transmission, limiting the functionality of these devices. Photoplethysmography (PPG)-based wearables, such as smartwatches and heart rate monitors, are the most prevalent for physiological monitoring in swimmers; however, their accuracy and precision when positioned on the wrist have been found to be reduced during freestyle swimming due to motion artifacts and suboptimal skin contact [13]. While Inertial Measurement Units (IMUs) are capable of measuring and analyzing stroke kinematics, their reliability is inferior to that of traditional video analysis [14]. Smart goggles with integrated motion sensors enable the real-time display of pacing and stroke characteristics; yet, they underscore the gradual pace of innovation in this domain of sports monitoring.

Near-infrared spectroscopy (NIRS)-based wearable sensors offer a novel means of monitoring key physiological biomarkers relevant to human performance [15]. Muscle Oxygen Saturation (SmO₂) reflects the balance between oxygen delivery and utilization within a muscle, and its continuous monitoring has the capability to provide unique insights into the oxidative metabolism of the target muscle and, consequently, its muscular endurance [16]. While NIRS SmO₂ monitoring is becoming more widespread within sports performance settings, it is currently underutilized in the world of swimming [17]. A few previous studies have begun to explore SmO₂ within the context of swimming. Pratama and Yimlamai measured the SmO₂ of the biceps femoris during three different rest protocols in collegiate swimmers [18]. Similarly, Dalamitros et al. utilized SmO₂ measurement on the middle deltoid to evaluate the acute responses to interval sprint training in national-level swimmers [19]. However, muscle oxygenation was measured outside the water immediately following the swims in both studies. To our knowledge, only one other study by Jones and Cooper has continuously measured SmO₂ during a swim [20]. In this study, SmO₂ was simultaneously measured on the subjects’ vastus lateralis and latissimus dorsi, comparing club-level swimmers against club-level triathletes during a 200 m freestyle swim.

This study aims to evaluate the efficacy of a five-week functional exercise intervention via a randomized control trial (RCT) on enhancing the muscular endurance of the posterior shoulder musculature in collegiate butterfly swimmers. We hypothesize that the intervention will induce significant adaptations in shoulder endurance as measured by intra-swim SmO₂ characteristics and improvements in the Posterior Shoulder Endurance Test (PSET), a metric inversely correlated with shoulder pain in swimmers [21]. This is the first study to examine how dryland resistance training translates to SmO₂ characteristics during swimming. Additionally, this study represents the first RCT in assessing the role of wearable NIR technology to provide objective data to demonstrate improvements in an exercise toward optimizing human performance. As such, this RCT fills a major void in the current orthopedic sports medicine field and presents the next steps at the nexus of electronics and sports medicine toward engineering, validating, and translating technologies to improve health and performance.

2. Materials and Methods

A dryland exercise was implemented to mimic the hand path and resistance profile of the recovery phase. The objective of this exercise was to enhance the muscular endurance of the posterior shoulder musculature, with the ultimate goal of reducing shoulder fatigue during the recovery phase.

2.1. Participants

Twelve (n = 12) Division I collegiate swimmers were recruited and provided written consent for this study (n = 4 female, n = 8 male). Individual athlete characteristics are displayed in

Table 1. Assorted Participant Data.

Athlete	Group	Sex	Age (yrs)	Height (cm)	Weight (kg)	BMI	Years Swam	World Aquatics Points
Athlete 1	Exercise	M	21	180	83.9	25.8	16	713.65
Athlete 2	Exercise	M	20	183	87.1	26	10	713.2
Athlete 3	Exercise	F	21	172	74.8	25.1	15	712.65
Athlete 4	Exercise	M	19	178	67.1	21.2	14	663.5
Athlete 5	Exercise	F	20	172	68	22.8	12	703.45
Athlete 6	Exercise	M	20	191	83.9	23.1	11	725.7
Athlete 7	Control	M	20	188	79.4	22.5	13	630.5
Athlete 8	Control	M	20	178	72.6	23	10	638.2
Athlete 9	Control	F	19	183	72.6	21.7	11	730.15
Athlete 10	Control	M	22	175	74.8	24.4	15	588.4
Athlete 11	Control	F	20	175	64.9	21.1	12	636.55
Athlete 12	Control	M	20	190	83.9	23.1	15	697.35
Averages:	6 EG; 6 CG	8M; 4F	20.17 ± 0.83	180.54 ± 6.45	76.25 ± 7.49	23.32 ± 1.67	12.83 ± 2.12	697.44 ± 46.22

. Participants were paired based on sex and randomized to either the exercise group (EX) or the control group (CTRL). This study was conducted in accordance with the Lehigh IRB (2113291-4). The inclusion criteria for this study were as follows: 1. Participants must be 18 years of age or older; 2. Participants must be academically eligible as per the following criteria set forth by the Athletic Department at Lehigh University; 3. Athletes must compete in the butterfly event.

2.2. Training Procedure

Before the onset of this study, all participants performed the Posterior Shoulder Endurance Test (PSET) using their right arm with a load of 2% body weight, as outlined by Moore et al. [22]. Participants synchronized their movements to a metronome, maintaining a steady pace of 30 beats per minute during the test. This test was performed again at the conclusion of this study. All subjects participated as normal in the team practices and dryland sessions. Once per week, athletes performed a 100-yard butterfly sprint, equipped with the MOXY sensor on the subject’s posterior deltoid. An overview of this study’s timeline is presented in Figure 1a.

For the CTRL group, participants only engaged in their regular team practices and dryland sessions without any additional interventions. In contrast, the EX-group underwent a specific exercise intervention designed to enhance posterior shoulder endurance in addition to the regular team practices and dryland sessions. For the EX group, a familiarization session was provided to instruct the participants on the correct exercise technique and to determine the starting load. The exercise intervention consisted of 3 weekly sessions, with at least 48 h between sessions. Each session consisted of two sets of exercises performed until muscular failure. A 3-min rest period was given between sets. The weight was incrementally increased once the participant was able to perform the exercise, without form breakdown, for 60 s. Subjects were supervised by the investigators during all training sessions.

2.3. Exercise Description

Participants laid chest-down on an exercise bench set at an incline of 15 degrees, holding dumbbells in each hand. The starting position began with the subject holding dumbbells in their hands, with their arms perpendicular to the floor. The participant then extended their arms backward until the arm was parallel to their torso. From this position, the subject abducted their arms until they were perpendicular to their torso and parallel to the floor. Then, the subject lowered their arms back down to the starting position in a controlled manner. At each of these positions, the subject was instructed to pause briefly to minimize momentum throughout the movement. The three positions are depicted in Figure 1b.

To maximize muscular adaptations, the following progressive overload scheme was employed. The weight used for the exercise was incrementally increased by 1 lb per arm once the athlete was able to perform the exercise without form breakdown for 60 s. Any one of the following conditions was determined to be indicative of form breakdown: inability to hold the brief pause at any of the three positions; excessive rest at the bottom of the movement (>1 s); compensation by elevation of the torso; or a significant amount of scapular retraction. It is important to note that while these criteria are used for the determination of form breakdown, this was solely for the purpose of weight increases. Form breakdown did not mark the termination of the exercise, as it was performed until complete muscular failure.

2.4. Data Collection

SmO₂ was assessed using a wireless NIRS sensor (Moxy Monitor, Fortiori Design LLC., Hutchinson, MN, USA). The sensor was placed at the midpoint of the posterior deltoid, 2 cm below the acromion’s angle, following the electrode placement guidelines recommended by SENIAM [23]. The device’s orientation was adjusted so that the sensor diodes were parallel with the direction of the posterior deltoid fibers. To ensure secure attachment to the subjects, a double-sided adhesive was employed at the bottom of the device, supplemented with waterproof athletic tape to improve stability, block out ambient light, and enhance subject comfort without impeding the range of motion. The sampling frequency of the sensor was set to collect data at 1 Hz, with data smoothing enabled.

2.5. Statistical Analysis

Dual linear regression was conducted for all weeks to determine the point of desaturation and r-values. Dual linear regression was chosen based on the work by Batterson et al., which observed consistent desaturation and sustained states during exercise in longer monitoring durations [24]. The point of desaturation was identified by the point of inflection (POI). If the POI exceeded 25 s and the r² value was below 0.5, desaturation was not observed. r² values were interpreted as follows: >0.9 indicated a very strong relationship; >0.7 indicated a strong relationship; and >0.5 indicated a moderate relationship. A one-sided t-test with an alpha of 0.05 was conducted for all athletes between week one and week five. For weeks where the dual regression r-values were higher in week two or week four, a one-sided t-test was performed for week one and week four or week two and week five. The significance of the p-values was denoted with asterisks: <0.01 as very strong **; <0.05 as strong *; <0.1 as moderate; and <0.5 as weak. If the r-value was weak, no significant POI was found.

3. Results

Forty-eight of the 51 swims measured reached at least a moderate r-value (r² > 0.5). Of these, 13 showed an r-value that represented a very strong relationship (r² > 0.9). These findings validate the use of dual linear regression as a mechanism to find the point of desaturation. In conjunction with the r-values found, observing the comparative line graphs in Figure 2 and Figure 3 reveals this dual linear pattern. Despite the shorter monitoring period and more significant variance in the data contributing to lower r² values, 94% of the data exhibited at least a moderate relationship, supporting the consistency of the dual linear shape observed in our findings.

Figure 2. Comparative SmO₂ Graphs for Week 1 and Week 5. (a–f) correspond to the exercise group (Athletes 1–6); (g–l) correspond to the control group (Athletes 7–12).

Figure 2. Comparative SmO₂ Graphs for Week 1 and Week 5. (a–f) correspond to the exercise group (Athletes 1–6); (g–l) correspond to the control group (Athletes 7–12).

Figure 3. Comparative graphs for the selected weeks. Chosen weeks were determined based on the POI r² values found in Table 2 and Table 3. (a–f) correspond to the exercise group (Athletes 1–6); (g–l) correspond to the control group (Athletes 7–12).

Figure 3. Comparative graphs for the selected weeks. Chosen weeks were determined based on the POI r² values found in Table 2 and Table 3. (a–f) correspond to the exercise group (Athletes 1–6); (g–l) correspond to the control group (Athletes 7–12).

Table 2. EX Group Point of Inflection and Desaturation Rates.

		Athlete 1	Athlete 2	Athlete 3	Athlete 4	Athlete 5	Athlete 6
Week 1	POI (s)	27.28 ± 2.26	9.49 ± 2.93	16.33 ± 0.47	40.60 ± 2.00	49.00 ± 21.87	9.53 ± 0.40
	POI R2	0.6691	0.872	0.9638	0.481	0.0059	0.9289
	Desat. Slope (SmO2/s)	−2.61	−4.85	−2.58	N/A	N/A	−5.76
	Desat. R2	0.7118	0.9641	0.9129	N/A	N/A	0.9807
Week 2	POI (s)	8.36 ± 0.63	6.53 ± 0.60	22 ± 1.19	5.55 ± 0.51	47.40 ± 3.04	19.53 ± 0.79
	POI R2	0.8071	0.7503	0.8957	0.7361	0.4183	0.9012
	Desat. Slope (SmO2/s)	−5.7	−7.77	−3.22	−6.18	N/A	−4.81
	Desat. R2	0.938	0.8847	0.9412	0.9699	N/A	0.8192
Week 3	POI (s)	19.46 ± 2.26	20.626 ± 1.74	26.53 ± 1.47	10.79 ± 0.68
	POI R2	0.5691	0.7482	0.8734	0.8324
	Desat. Slope (SmO2/s)	−4.7	−2.44	−2.08	−4.91
	Desat. R2	0.6856	0.8451	0.7684	0.9791
Week 4	POI (s)	16.01 ± 0.80	7.68 ± 0.59	10.42 ± 0.48	10.42 ± 0.48	20.42 ± 1.13	14.31 ± 0.35
	POI R2	0.9416	0.8135	0.9025	0.9025	0.885	0.9778
	Desat. Slope (SmO2/s)	−3.44	−6.65	−5.74	−5.37	−2.6	−3.84
	Desat. R2	0.9758	0.9349	0.9824	0.9893	0.832	0.9535
Week 5	POI (s)	16.76 ± 0.71	28.47 ± 1.48	12.81 ± 0.76	15.74 ± 0.77	21.00 ± 1.68	12.34 ± 0.61
	POI R2	0.9574	0.7191	0.8497	0.9205	0.7023	0.8807
	Desat. Slope (SmO2/s)	−4.71	−2.19	−4.73	−4.85	−2.79	−6.05
	Desat. R2	0.9555	0.7786	0.9875	0.8037	0.8808	0.9261

Boxes for which data are not present represent instances when the athlete was not present and did not complete the swim.

Table 2. EX Group Point of Inflection and Desaturation Rates.

		Athlete 1	Athlete 2	Athlete 3	Athlete 4	Athlete 5	Athlete 6
Week 1	POI (s)	27.28 ± 2.26	9.49 ± 2.93	16.33 ± 0.47	40.60 ± 2.00	49.00 ± 21.87	9.53 ± 0.40
	POI R2	0.6691	0.872	0.9638	0.481	0.0059	0.9289
	Desat. Slope (SmO2/s)	−2.61	−4.85	−2.58	N/A	N/A	−5.76
	Desat. R2	0.7118	0.9641	0.9129	N/A	N/A	0.9807
Week 2	POI (s)	8.36 ± 0.63	6.53 ± 0.60	22 ± 1.19	5.55 ± 0.51	47.40 ± 3.04	19.53 ± 0.79
	POI R2	0.8071	0.7503	0.8957	0.7361	0.4183	0.9012
	Desat. Slope (SmO2/s)	−5.7	−7.77	−3.22	−6.18	N/A	−4.81
	Desat. R2	0.938	0.8847	0.9412	0.9699	N/A	0.8192
Week 3	POI (s)	19.46 ± 2.26	20.626 ± 1.74	26.53 ± 1.47	10.79 ± 0.68
	POI R2	0.5691	0.7482	0.8734	0.8324
	Desat. Slope (SmO2/s)	−4.7	−2.44	−2.08	−4.91
	Desat. R2	0.6856	0.8451	0.7684	0.9791
Week 4	POI (s)	16.01 ± 0.80	7.68 ± 0.59	10.42 ± 0.48	10.42 ± 0.48	20.42 ± 1.13	14.31 ± 0.35
	POI R2	0.9416	0.8135	0.9025	0.9025	0.885	0.9778
	Desat. Slope (SmO2/s)	−3.44	−6.65	−5.74	−5.37	−2.6	−3.84
	Desat. R2	0.9758	0.9349	0.9824	0.9893	0.832	0.9535
Week 5	POI (s)	16.76 ± 0.71	28.47 ± 1.48	12.81 ± 0.76	15.74 ± 0.77	21.00 ± 1.68	12.34 ± 0.61
	POI R2	0.9574	0.7191	0.8497	0.9205	0.7023	0.8807
	Desat. Slope (SmO2/s)	−4.71	−2.19	−4.73	−4.85	−2.79	−6.05
	Desat. R2	0.9555	0.7786	0.9875	0.8037	0.8808	0.9261

Boxes for which data are not present represent instances when the athlete was not present and did not complete the swim.

Table 3. CTRL Group Point of Inflection and Desaturation Rates.

		Athlete 7	Athlete 8	Athlete 9	Athlete 10	Athlete 11	Athlete 12
Week 1	POI (s)	9.46 ± 0.37	24.85 ± 3.90	5.10 ± 0.69	17.74 ± 0.76	7.65 ± 0.40	7.62 ± 0.40
	POI R2	0.9398	0.5461	0.5292	0.8788	0.8854	0.888
	Desat. Slope (SmO2/s)	−5.45	−1.69	−5.86	−2.99	−4.62	−4.83
	Desat. R2	0.9947	0.7898	0.9796	0.9399	0.9452	0.96
Week 2	POI (s)		18.95 ± 1.20	8.53 ± 0.52			14.76 ± 0.61
	POI R2		0.8592	0.8333			0.9284
	Desat. Slope (SmO2/s)		−3.25	−6.4			−4.52
	Desat. R2		0.8955	0.9754			0.9512
Week 3	POI (s)	10.27 ± 0.52		49.49 ± 2.23	44.38 ± 1.80	24 ± 1.46
	POI R2	0.97		0.6654	0.6527	0.86
	Desat. Slope (SmO2/s)	−4.65		N/A	N/A	−1.75
	Desat. R2	0.985		N/A	N/A	0.7296
Week 4	POI (s)		9.50 ± 0.37	20 ± 3.01		12.01 ± 0.94	9.70 ± 2.02
	POI R2		0.9444	0.4864		0.7639	0.3121
	Desat. Slope (SmO2/s)		−4	−3.74		−3.03	−5.52
	Desat. R2		0.9182	0.6805		0.9922	0.9279
Week 5	POI (s)	7.38 ± 0.37	21 ± 2.19	12.28 ± 0.73	9 ± 0.98	20 ± 0.91	9 ± 1.51
	POI R2	0.8886	0.7732	0.8266	0.6363	0.9016	0.6641
	Desat. Slope (SmO2/s)	−4.7	−2.79	−5.56	−4.28	−2.83	−4.75
	Desat. R2	0.9665	0.8079	0.8384	0.9795	0.9409	0.9183

Boxes for which data are not present represent instances when the athlete was not present and did not complete the swim.

Table 3. CTRL Group Point of Inflection and Desaturation Rates.

		Athlete 7	Athlete 8	Athlete 9	Athlete 10	Athlete 11	Athlete 12
Week 1	POI (s)	9.46 ± 0.37	24.85 ± 3.90	5.10 ± 0.69	17.74 ± 0.76	7.65 ± 0.40	7.62 ± 0.40
	POI R2	0.9398	0.5461	0.5292	0.8788	0.8854	0.888
	Desat. Slope (SmO2/s)	−5.45	−1.69	−5.86	−2.99	−4.62	−4.83
	Desat. R2	0.9947	0.7898	0.9796	0.9399	0.9452	0.96
Week 2	POI (s)		18.95 ± 1.20	8.53 ± 0.52			14.76 ± 0.61
	POI R2		0.8592	0.8333			0.9284
	Desat. Slope (SmO2/s)		−3.25	−6.4			−4.52
	Desat. R2		0.8955	0.9754			0.9512
Week 3	POI (s)	10.27 ± 0.52		49.49 ± 2.23	44.38 ± 1.80	24 ± 1.46
	POI R2	0.97		0.6654	0.6527	0.86
	Desat. Slope (SmO2/s)	−4.65		N/A	N/A	−1.75
	Desat. R2	0.985		N/A	N/A	0.7296
Week 4	POI (s)		9.50 ± 0.37	20 ± 3.01		12.01 ± 0.94	9.70 ± 2.02
	POI R2		0.9444	0.4864		0.7639	0.3121
	Desat. Slope (SmO2/s)		−4	−3.74		−3.03	−5.52
	Desat. R2		0.9182	0.6805		0.9922	0.9279
Week 5	POI (s)	7.38 ± 0.37	21 ± 2.19	12.28 ± 0.73	9 ± 0.98	20 ± 0.91	9 ± 1.51
	POI R2	0.8886	0.7732	0.8266	0.6363	0.9016	0.6641
	Desat. Slope (SmO2/s)	−4.7	−2.79	−5.56	−4.28	−2.83	−4.75
	Desat. R2	0.9665	0.8079	0.8384	0.9795	0.9409	0.9183

Boxes for which data are not present represent instances when the athlete was not present and did not complete the swim.

The box and whisker plots provide insights into the overall noise during the sustained monitoring period. Larger ranges in the plots correlated with lower r² values, while smaller ranges were associated with higher r² values. For the three weeks that did not follow the hypothesized dual linear relationship, the entire dataset was used since a singular point of desaturation, and a district-sustained region could not be identified. In the exercise group, the sustained SmO₂ levels showed a negative trend, with the median levels displaying a clear downward trend.

A one-sided t-test with an alpha of 0.05 was conducted for all athletes between week one and week five, testing if week one values were greater than week five values (

Table 4. P-values for the significance between weeks, as determined by 1-sided t-test.

	Week 1 vs. Week 5	Chosen Weeks
Athlete 1	0.7314	0.4194
Athlete 2	0.9835	0.009062
Athlete 3	3.85 × 10−8	1.67 × 10−7
Athlete 4	2.38 × 10−5	0.0001241
Athlete 5	4.08 × 10−13	2.20 × 10−16
Athlete 6	0.9796	0.9583
Athlete 7	0.7446	0.7446
Athlete 8	0.9724	0.9496
Athlete 9	0.2516	1
Athlete 10	1	1
Athlete 11	1	1
Athlete 12	0.5542	0.4325

). For weeks where the dual regression r-values were higher in week two or week four, a one-sided t-test was performed for week one and week four or week two and week five. Athletes 2 through 5 all presented very strong p-values for their t-tests, in which the weeks chosen were based on the r-values. Athletes 3 through 5 maintained these very strong p-values when comparing only their week 1 and week 5 data. In contrast, athletes 7 through 12 had no statistically significant p-values (p < 0.05) for both the r-value chosen weeks and for week 1 and week 5.

The exercise group showed a statistically significant increase (p < 0.05) in PSET using a difference in means one-way t-test, while the control group did not show this statistically significant increase (p < 0.05) (

Table 5. PSET scores for both exercise (EX) and control (CTRL) groups.

	EX	CTRL
PRE	31 ± 4.9	34 ± 7.6
POST	67.7 ± 15.7	40.5 ± 11.8
p-value	0.0014	0.154

). The baseline scores for the PSET were relatively similar between the two groups. However, after the duration of this study, the average score of the EX-group increased over two-fold, while the CTRL group did not see statistical significance (p < 0.05). Slight but notable hypertrophy in several of the swimmers’ posterior deltoids was also observed, although no procedure was set to quantify this.

4. Discussion

Swimmers in the EX-group experienced several statistically significant (p < 0.05) adaptations due to the exercise intervention. Firstly, while both groups increased their PSET scores, only EX’s increase was statistically significant (p < 0.01). These results are in alignment with the principle of specificity since the movement progression from Position 3 to Position 1 during the exercise closely follows the eccentric phase of the PSET movement. Every athlete in the CTRL group increased their PSET score, although not to a significant degree (p > 0.05), which was likely caused by reconditioning due to the commencement of this study after a detraining period. This increase in PSET score is likely to indicate an increase in the muscular endurance of the posterior shoulder musculature during external rotation. This factor has been implicated to play a significant role in pain reduction and injury prevention [21,22]. A prospective study on youth swimmers by Feijin et al. revealed that for every one rep increase in PSET, the odds for shoulder pain decreased by 5% [21]. McKenzie et al. proposed a mechanism for this, such that increasing the endurance of the posterior rotator cuff could help stabilize the humeral head in the glenoid while swimming, therefore reducing the risk of functional impingement [25]. This exercise proves to be a viable intervention for improving the muscular endurance of the posterior shoulder musculature, as measured by PSET.

The increase in muscular endurance was also quantified through the muscle oxygenation profiles of the swimmers during 100-yard butterfly sprints. Looking at the sustained SmO₂ levels during the swims (Figure 4), a negative trend can be observed for the EX, but no consistent pattern can be observed in the CTRL group. Sustained levels of SmO₂, on a cellular level, indicate that the muscle can extract oxygen from the blood at a sufficient rate to maintain its current metabolic demands. When the sustained level of SmO₂ is decreased, the oxidative capacity of the muscle increases, and it can utilize more available oxygen, increasing fatigue resistance [26]. The adaptations seen in the SmO₂ profiles likely reflect an alteration in the onset of fatigue, an effect that might not be significant, given that 100-yard butterfly sprints are not the best event for measuring endurance. The increase in muscular endurance would likely carry over well to the freestyle stroke as well, where Matthews et al. described how fatigue in the rotator cuff muscles resulted in a decrease in external rotation while swimming [27]. It is also likely that this increase in muscular endurance would result in improvements in the swimming rhythm, as the decreases in stroke length and efficiency would occur later in the swim due to the offset of fatigue [2].

A few studies have demonstrated the relationship between muscle activation and muscle oxygenation [28]. No definitive conclusions about the exercise and its effects on posterior shoulder activation can be drawn since EMG data were not recorded. It is hypothesized that the exercise intervention induced greater activation of the posterior deltoid for Athlete 5 during the 100-yard butterfly swim. Initial exercise sessions revealed a trapezius-dominant movement pattern, slowly transitioning to a more balanced movement pattern throughout the intervention. Weeks 1 and 2 displayed no desaturation in the posterior deltoid, with the graph remaining relatively constant over the duration of the swim, implicating that the muscle was not being activated. In weeks 4 and 5, the SmO₂ graph displayed the expected Desaturation-Sustained pattern (r² > 0.70). A possible mechanism for this activation could be a shift in motor recruitment patterns of the surrounding musculature. Although not directly adjacent, previous research has demonstrated that targeted isolation exercises could modify recruitment patterns of the primary agonists in a dynamic movement, such as those involved in the bench press [29]. While no EMG data were collected in this experiment, Brown et al. found that an isometric hold similar to Position 2 in Figure 1 elicited significant activation of the posterior deltoid [30]. These findings indicate that the exercise intervention may have influenced muscle recruitment in a way that enhances posterior deltoid engagement during the butterfly stroke.

Based on experimental observations, it is hypothesized that shoulder fatigue during the recovery phase could be a factor in the incidence of lower back pain prevalent in butterfly swimmers [31,32]. It was observed that fatigue in the exercise resulted in several participants inducing spinal flexion by lifting their torso as a compensatory mechanism for shoulder fatigue. Although not in swimmers, Dupois et al. discovered that induced shoulder fatigue led to alterations in the kinematics of both the shoulder and trunk, resulting in an observed increase in spinal flexion [33]. De Jesus et al. describe how the percentage of the stroke cycle spent in the recovery phase is greater during the fourth lap when compared to the first lap [4]. The vertical foot amplitude on the second downbeat kick, which occurs at the beginning of the recovery phase, also follows this same trend [4]. The combination of elevating the torso while kicking through a larger range of motion likely places the spine in an unfavorable arched position, which could result in lower back pain when repeated over many stroke cycles.

The present study is characterized by several notable limitations. Firstly, the placement of the device on the posterior deltoid is susceptible to significant dynamic movement, which presents challenges in maintaining adhesion and increases the likelihood of noise artifacts. Observed noise in the data, likely from poor adhesion or improper sensor placement, has led to low R² values, especially during weeks 1 and 2. This noise coincides with elevated median SmO₂ levels, causing the data to not be normally distributed. If ANOVA, the most appropriate test for this study, had been applied, it could have resulted in an inflated representation of SmO₂ improvements across the five weeks. The exercise itself would have ideally been conducted on a flat bench with a 0-degree angle to best mimic the body positioning of the butterfly stroke. However, the 15-degree angle was implemented to grant adequate vertical space for all athletes to conduct the exercise using the full range of motion while minimizing the deviation from optimal butterfly body positioning. While five athletes in the EX-group missed one or less of the exercise sessions, Athlete 6 only attended one or two sessions per week due to outside conflicts. Interestingly, while no significant adaptations in muscle oxygenation were observed, Athlete 6 still increased their PSET score from 24 to 42, a 75% increase. This suggests that the high-frequency volume of the exercise may not be required to improve external rotation endurance. While these results are promising, a small sample size and an incomplete dataset require further investigation to validate these findings.

5. Conclusions

This is the first randomized controlled trial to assess the role of wearable NIRS sensors for optimizing swimming training. In a population of collegiate butterfly swimmers, this work demonstrates how dryland exercise translates to adaptations in the water, as measured by SmO₂. The exercise introduced in this paper induced significant adaptations in muscular endurance. PSET performance was significantly different, increasing by an average of 36 in the experimental group but only 6 in the CTRL group. Four out of the six athletes in the EX-group significantly lowered their SmO₂ over the duration of a 100-yard butterfly sprint. This change was not observed in the CTRL group. Given the small sample size and incomplete dataset, further investigation is required to validate these findings. Future works should continue exploring the application of this movement in other contexts. Performance improvements could be observed in longer-distance events, such as the 200-yard butterfly, where the effects of muscular endurance will be more pronounced. Kinematic analysis could be applied to determine changes in stroke mechanics. Given the improvements in PSET, this exercise could also be applied in other settings or overhead sports where external rotation endurance is beneficial for injury prevention. These results not only highlight the potential for performance improvements but also emphasize the broader implications of sports medicine research in translating performance analytics into disease modeling and treatment frameworks.