R. rosea root extract, standardized to SA, was prepared according to a previously published protocol (Liu et al., 2021). Both extraction and drying were performed at Professor Yi-Ming Chen’s laboratory. The placebo comprised isocaloric corn-starch powder packed in identical capsules, rendering the two interventions visually indistinguishable to both participants and investigators.

An a priori power analysis was performed in G*Power v3.1.9.7. Assuming a medium effect size (Cohen’s d = 0.50) for the primary outcome (time-to-exhaustion), α = 0.05 (two-tailed) and power = 0.80, a minimum of 24 participants per group was indicated. However, because the participant pool was restricted to students enrolled in a single physical-education course, we were able to recruit only 30 eligible recreationally active women, resulting in smaller group sizes than the a priori target. Participants were classified as recreationally active based on a priori criteria. All volunteers completed the Physical Activity Questionnaire-Short Form (IPAQ-SF, and eligibility required ≥3 sessions·week⁻¹ of structured moderate-to-vigorous activity (≥30 min per session; Borg RPE ≥12) over the preceding 3 months, with no involvement in varsity/elite or professionally supervised training. The sample reported [1605 ± 332] MET·min·week⁻¹. Common activities included jogging, group aerobic classes, recreational ball games, and cycling for transport. None had prior experience with flywheel-based or formal plyometric programs. This level of activity aligns with a Division III/club sport profile habitually active but not undergoing high-volume, periodized conditioning.

Utilizing a randomized, double-blind, parallel-group design, this study included a total of 30 healthy female college students enrolled in a physical education course and randomly allocated them to one of three groups: a sedentary control group that received supplementation with dextrose placebo powder (SC), the eccentric exercise group that received supplementation with placebo (ECC), and the eccentric exercise group that received supplementation with SA (ECC + SA). All three groups showed no significant difference in age, height and weight (age = 19.0 ± 1.4 years; height = 165.0 ± 3.9, 164.6 ± 6.8, and 164.7 ± 4.8 cm; weight = 56.8 ± 6.0, 56.9 ± 10.3, and 57.0 ± 10.3 kg, respectively). None of the participants had any record of cardiovascular or respiratory disease or musculoskeletal injury. Participants who had trained in any jumping program were excluded. All study procedures were approved by the Joint Institutional Review Board of Chinese Clinical Trial Register (ChiCTR1900027683). This study was conducted in accordance with the ethical principles stated in the Declaration of Helsinki. After enrollment, both SA and placebo powders were encapsulated in soft pine bark capsules. Each group (n = 10) consumed 150 mg/day of SA or placebo supplements before breakfast for 4 weeks. The ECC and ECC + SA groups came to our laboratory every day for eccentric exercise assessment, with both groups performing the same number of sets at the same intensity (i.e., percentage body weight). Besides ECC training, all subjects were asked to maintain normal activity and refrain from any strenuous exercise. Subjects participate 2 study visits on Day 0 (pretest) and Day 28 (posttest). After the 4-week supplementation, a posttest was conducted. At posttest, participants guessed capsule identity (SC vs. SA); correct identification did not differ from chance (χ² = 0.27, p = 0.60), confirming successful blinding.

Participants were instructed to maintain habitual intake at approximately 2,200–2,300 kcal·day^-1, based on an estimated activity level of 1.6–1.7 (1650 MET·min·wk⁻¹). The target macronutrient distribution was 55% carbohydrates, 20% protein, and 25% fat.”

At the beginning and end of the experiments, the body weight (kg), body height (cm), body mass index (BMI; kg/m²) fat-free mass (FFM, kg), body fat percentage (% of body weight), and water content percentage (%) of all 30 participants were measured using a body fat analyzer (MC-980, Tanita Co., Tokyo, Japan) based on a previously reported method (Chen et al., 2021).

Eccentric exercise training was performed using the flywheel equipment (K-box4, Exxentric ABTM, Bromma, Sweden). Before the training period, all participants completed the training sessions to familiarize themselves with the training process, including a review of the safety criteria and practice with squat training using a flywheel gravity device with an inertia of 0.01 kg/m². The tension band of the K-box4 device was adjusted according to the height of individual participants, with the squat range of motion being standardized for all participants. The length of the tension band was used to determine the range of motion of the knee joint angle for all subjects. Before starting the training program, all participants performed a 10-min warm-up. During the training program, participants performed 15 repetition maximum of deep flywheel squats with a speed of 0.01 kg/m² inertia load. The range of motion of the participants was standardized, and a tape was used to ensure that the knee angle reached 90 degrees (thighs parallel to the ground). Consequently, the participants performed squats by extending their knees (180 degrees) from a low (90 degrees) position. Ankle extension was not allowed. Participants were instructed to perform the concentric action phase as quickly as possible while delaying braking until the last part of the eccentric phase. Throughout the training program, the participants were given loud verbal encouragement. The training was designed to overload the eccentric phase using the flywheel (K-Box4). Although a brief concentric action is required to accelerate the wheel, the primary stimulus was the deceleration phase, during which participants were instructed to delay braking to maximize eccentric loading.

The exhaustive cycling test was performed at the 26th day of the intervention. To avoid the effects of dietary intake on outcome measures during testing days, subjects ate a standardized breakfast on the morning of the test day. The standard breakfast comprised 55% carbohydrates, 15% protein, and 30% lipids, for a total caloric intake of 380 kcal. Tests were started 30 min after all participants consumed the supplement 30 min. Exhaustive tests were performed using a face mask on a cardiopulmonary ergospirometry device. Subjects were instructed to avert any strenuous exercise activity for 3 days before the biking exercise test for VO₂max analysis. The speed and wattage of the bike were based on the ramp protocol on a cycle (Michalik et al., 2019). VO₂ and VCO₂ were measured breath by breath via calibrated metabolic cart (Schiller CS200, Schiller AG, Switzerland). The ramp protocol increased workload by 15 W min⁻¹ until volitional exhaustion. VO₂ max was defined as the highest 30s averaged VO₂ accompanied by at least two of the following: RER ≥1.10, HR ≥ 90% of age-predicted maximum, or a VO₂ plateau (<150 mL min⁻¹ increase despite workload increment). After the exhaustive endurance exercise challenge, lactate, ammonia, glucose, CK, lactate dehydrogenase (LDH), and free fatty acid (FFA) levels were assessed using a Beckman Coulter AU5800 autoanalyzer (Beckman Coulter Inc., Brea, CA, United States). At the end of the experiment (30 days), all subjects had fasted for at least 8 h and took the last supplementation dose, after which serum biochemical indices were assessed using a Beckman Coulter AU5800 autoanalyzer (Beckman Coulter Inc., Brea, CA, United States).

The total number of 160 jumps was selected based on protocols used in prior fatigue studies aiming to observe longitudinal decrements in explosive performance and stretch-shortening cycle function. Although this setup does not mimic a single real-game scenario, it maintains ecological validity by reproducing the intensity and density of jump tasks common in sport-specific training and match play. Additionally, previous studies investigating the effects of nutritional or ergogenic supplementation on neuromuscular performance have also adopted similar high volume DJ protocols (160 jumps) to reliably induce fatigue and assess intervention efficacy (Chen et al., 2021). All participants wore standardized shoes and shorts and completed a familiarization session before data collection. After a dynamic warm-up (5 min of treadmill running at 6 km h^-1 plus lower-limb stretches), the participants rested for 5 min and then performed three countermovement jumps (CMJs) on dual force plates. During each CMJ, the participant began from an upright stance, descended to ∼90° knee flexion, and finished with a maximal vertical takeoff while maintaining arm swing, with each attempt separated by 60 s of passive recovery. The highest CMJ was defined as 100% jump height (H). Target drop-jump heights were set at 70%, 100%, and 130% of H. Notably, 130% H was selected as the test height after 3 min of rest and five familiarization trials. Subjects then completed 160 drop vertical jumps at this height with 10 s between jumps, subsequently landing and rebounding on the force plates with the hands placed on the waist to eliminate arm assistance. Data processing was performed using a custom MATLAB program (version 2019a, MathWorks Inc., Natick, MA, United States). A fourth-order low-pass Butterworth digital filter with a cutoff frequency of 50 Hz can smooth the ground reaction forces and marker trajectories. The intervention effects were assess using the parameters of the reactive strength index, which was calculated using the following formula (Ward et al., 2019; Wang et al., 2021): Reactive strength index  RSI = JH Contact   time , where contact time was defined as the duration from initial contact, at which the vertical ground reaction force exceeded a threshold of 10 N, to toe-off in an instant. RSI can represent the response speed of the participants from touchdown to takeoff.

Statistical analyses were performed using MATLAB software (version 2019a, MathWorks Inc., Natick, MA, United States). All data were expressed as means ± SD. One-way analysis of variance (ANOVA) followed by post hoc Tukey’s HSD tests was performed to determine significant differences in biochemistry and exhaustive exercise tests results between multiple groups. A mixed-design two-way ANOVA, with supplementation and time as independent variables, was used to compare pretest and posttest variables. A post hoc independent repeated t-test was used to determine the significance of the main and interaction effects. For all tests, a p value of <0.05 was considered significant. To assess the magnitude of differences, effect sizes were calculated using Cohen’s d. For between-group comparisons, Cohen’s d was computed based on pooled standard deviations. The interpretation of d was as follows: small (0.2 ≤ d < 0.5), medium (0.5 ≤ d < 0.8), and large (d ≥ 0.8). Effect sizes were reported alongside p-values to provide additional insight into the practical significance of the findings.

Anthropometric measurements are summarized in Table 1. A total of 30 participants underwent body composition measurements on before (pretest, day 0) and after (posttest, day 28) the intervention. Participants in the placebo, ECC and ECC + SA groups had a height of 165.0 ± 3.9, 164.6 ± 6.8, and 164.7 ± 4.8 cm, respectively. No significant differences in height were observed between each group. Aside from height, pretest measurements of body weight, fat mass (FM), FFM, BMI (kg/m²), and water content did not differ between the groups. For posttest measurements, no significantly difference in body weight, FM, FFM, BMI, and water content were observed between the SC, ECC, and ECC + SA groups. Thus, these findings suggest that the 4-week ECC and ECC + SA interventions did not influence body composition. Focusing on motion analysis, biochemistry tests, and exercise performance test could yield more precise results.

Before ECC training or ECC + SA supplementation, endurance exercise test performance was measured through an exhaustive cycling pretest (Figure 1). After 4 weeks of ECC training or ECC + SA supplementation, endurance exercise test performance was measured using an exhaustive cycling test. As shown in Figure 1A, the SC, ECC, and ECC + SA groups had a time to exhaustion of 14.3 ± 5.8, 14.4 ± 3.8, and 14.1 ± 4.8 min, respectively, in the pretest. No significant differences in time to exhaustion was observed between each group at pretest. At the posttest, the SC, ECC, and ECC + SA group showed a time to exhaustion of 13.7 ± 2.7, 18.2 ± 2.8, and 19.7 ± 7.0 min, respectively. Notably, time to exhaustion was 24.95% (ES = 1.64, p = 0.0375) and 30.80% (ES = 1.13, p = 0.0069) higher in the ECC and ECC + SA groups than in the SC group in the posttest. Figure 1B shows the pretest and posttest results of the maximal oxygen consumption (VO₂max) test in the SC, ECC, and ECC + SA groups. No significant differences in both pretest and posttest results were observed between each group (p = 0.6857, pretest and p = 0.3009, posttest). However, both the initial and maximum heart rates during the exhaustive cycling exercise test did not significantly differ between each group in the pretest and posttest (p = 0.7147 pretest and p = 0.8630 posttest, respectively; p = 0.9996 pretest and p = 0.9847 posttest, respectively; Figures 1C,D). Thus, our results demonstrated that 4 weeks of ECC training and ECC + SA supplementation prolonged the exhaustion time but did not enhance the oxygen consumption compared to the SC group. Overall, SA supplementation could improve endurance exercise performance during the eccentric training period.

As shown in Figure 2, two-way ANOVA revealed significant group × time interaction (all p ≤ 0.006) for the RSI from DJ1 to DJs160. Compared to the pretest results, pairwise comparison showed that the RSI in the ECC and ECC + SA groups increased significantly by 41% to 64% (ES = 1.01–1.59) and 55% to 111% (ES = 1.28–3.87), respectively, at the posttest (all p ≤ 0.011) after DJ1–DJs160. However, no significant difference between the pre- and posttest RSI was observed in the SC group. At the posttest, the RSI in the ECC and ECC + SA groups increased significantly by 41% to 61% (ES = 1.39–2.27, all p ≤ 0.044) and 50% to 106% (ES = 1.77–3.64, all p ≤ 0.012), respectively, after DJ1–DJs160 compared to the SC group. Specifically, compared to the ECC group, the ECC + SA group showed a significant increase in RSI (33%, 29%, 24%, 24%, and 23%, respectively) after DJs120 (ES = 1.53, p = 0.014), DJs130 (ES = 1.13, p = 0.030), DJs140 (ES = 1.29, p = 0.045), DJs150 (ES = 1.10, p = 0.021), DJs160 (ES = 1.08, p = 0.031) at the posttest. Thus, our findings suggested that an eccentric training protocol can increase the reactive strength index, but SA supplementation yielded an additional statistically significant gain that was most pronounced at the highest cumulative plyometric loads, implying that SA enhances explosive stretch–shortening cycle performance.

Biochemical analysis was used to determine the capacity for fatigue recovery in the SC, ECC, and ECC + SA group after the cycling test (Figure 3). Tests were conducted at three points: before and after acute exhausted cycling test (data before the cycling test not shown, no significant difference was found between each group, p > 0.05) and then after 90 min of rest after the cycling. In Figure 3A, lactate levels after exercise were 6.02 ± 2.08, 6.84 ± 2.60 and 5.72 ± 1.64 mmol/L in the SC, ECC, and ECC + SA groups, respectively. After 90 min of rest, lactate levels were 1.80 ± 0.72, 1.47 ± 0.66, and 1.56 ± 0.45 mmol/L in SC, ECC, and ECC + SA groups, respectively. No significant differences in lactate levels were observed between each group both after acute exercise and rest for 90 min (p = 0.4896 and 0.4776, respectively). Ammonia levels after exercise (Figure 3B) were 57.9 ± 26.9, 49.7 ± 16.7, and 46.2 ± 12.6 μmol/L in the SC, ECC, and ECC + SA groups, respectively. After 90 min of rest, ammonia levels were 17.4 ± 3.8, 20.5 ± 6.7, and 19.7 ± 9.7 (μmol/L) in the SC, ECC, and ECC + SA groups, respectively. No significant difference in ammonia levels were noted at both time points (p = 0.4162 and 0.6109, respectively). CK levels (Figure 3C) were 121 ± 27, 103 ± 30, and 89 ± 25 U/L in the SC, ECC, and ECC + SA groups, respectively, with significant differences having been observed after exercise (p = 0.0496). CK levels in the ECC + SA group decreased significantly by 26.20% (ES = 1.23, p = 0.0155) compared to that in SC group after exercise. After 90 min of rest, however, no significant differences in CK levels were observed between each group (p = 0.4362). As shown in Figure 3D, no significant differences in glucose levels were observed between each group after exercise (p = 0.3465) and 90 min of rest (p = 0.5716). Figure 3E revealed no significant differences in LDH levels at both time points (p = 0.2573 and 0.3194, respectively). FFA levels were 0.47 ± 0.17, 0.61 ± 0.21, and 0.68 ± 0.22 in the SC, ECC, and ECC + SA groups, respectively, after the cycling test. The ECC + SA group showed a significant 31.37 %increase in FFA levels (ES = 1.10, p = 0.0271) compared to the SC group after the cycling test. After 90 min of rest following the cycling test, no significant differences in FFA levels were observed between each group (p = 0.2032). These finding indicate that 4 weeks of ECC training could increase the time to exhaustion during cycling and may improve exercise economy but promoting increased fat utilization as a fuel source (FFA increase). Moreover, SA + ECC could reduce the muscle fatigue index (CK level) and enhance energy transport (FFA level) after endurance exercise.

After the 4-week ECC or ECC + SA intervention, health-related biochemical data were analyzed (Table 2). Notably, no significant differences in the pretest data were observed between each group (data not shown). Serum levels of AST, ALT, CK ALB, BUN, creatinine, glucose, total cholesterol (TC), and high-density lipoprotein (HDL) did not significantly differ between each group. Concerning the lipid profile, except for TC and HDL levels, TG levels decreased significantly by 36.82% (ES = 0.87, p = 0.0261) in the ECC + SA group compared to that in the SC group. Low-density lipoprotein (LDL) levels decreased significantly by 16.81% (ES = 0.9, p = 0.0295) and 15.10% (ES = 0.80, p = 0.0481) in both the ECC and ECC + SA groups compared to the SC group. Thus, our findings showed that both ECC training and SA supplementation had beneficial effects on the lipid profile.

This experiment was conducted exclusively in young, recreationally active women; hence, extrapolation to males or highly trained athletes awaits further verification. Although an a priori calculation indicated that 24 participants per group were required to achieve 80% power, logistical constraints limited enrolment to 10 per group, thereby reducing statistical power and external validity. To isolate the interaction between a fixed eccentric load and salidroside, fly-wheel inertia was intentionally kept constant; the absence of progressive overload, however, may have attenuated longer-term neuromuscular adaptations. Finally, because no oxidative-stress or inflammatory biomarkers were assessed, the mechanistic interpretation of salidroside’s antifatigue and muscle-protective effects remains speculative and should be confirmed in future work that incorporates biochemical and neuromuscular assays.