The effects of saddle height and power output on lower-limb muscles and joints during cycling Authors: Fangbo Bing, Tony Lin-Wei Chen, Guoxin Zhang, Yan Wang, Ming Zhang Published: 2026 ABSTRACT BackgroundCyclist numbers have risen sharply in the past decade, but overuse injuries are common. Proper saddle height and power output help prevent injuries by influencing riding posture and exercise intensity. This study examines how saddle height and power output affect lower limb loading, focusing on muscle forces and joint contact forces.MethodsTwenty-five participants cycled under nine riding conditions (three saddle heights × three power outputs). A musculoskeletal multibody model, incorporating detailed lower-limb muscles, was driven by marker trajectories and pedal forces and torques. Generalized estimating equations were used to assess the effects of saddle height and power output, adjusted for gender, BMI, and cadence. The mean muscle forces and the maximum joint forces during a cycle were examined. The degrees of association were estimated by the Wald statistic for a power output × saddle height interaction term. The outcomes for each cycling condition were compared against those of the null models. The significance was p < 0.05 with Holm‒Bonferroni correction for post hoc comparisons.ResultsThe predicted muscle activations were in good agreement with the electromyography data, which verified the model. Greater power outputs increased hip/knee extensor forces and joint forces (β = 0.021–0.065, p ≤ 0.026) but decreased ankle plantar flexion and joint forces (β = -0.041 – -0.038, p ≤ 0.019). Higher saddle heights decreased the flexion angles of the hip and knee and the dorsiflexion of the ankle. It further reduced the mean muscle forces (β = -0.355 – -0.096, p ≤ 0.026), including rectus femoris, vastus lateralis and medialis, gastrocnemius, and gluteus maximus, except for the tibialis anterior (β = 0.294, p = 0.015), and lowered most of the maximum directional joint forces (β = -0.064 – -0.126, p ≤ 0.045).ConclusionThe lower-limb joints were in a more flexed position at lower saddle heights, and higher power outputs and lower saddle heights increased kinetic load on lower-limb muscles and joints. Therefore, moderate power output with a higher saddle setting may be preferable. However, optimal adjustments should also consider other factors not examined in this study, such as saddle pressure.Supplementary InformationThe online version contains supplementary material available at 10.1186/s13102-026-01779-6. Introduction Cycling has exploded in popularity over the past decade. However, the repetitive nature of pedaling, combined with high training volumes, places cyclists at significant risk for overuse injuries, particularly in the lower limbs [1]. The knee is especially vulnerable, often linked to prolonged exposure to suboptimal biomechanical loads [2]. These loads are heavily influenced by riding power and bicycle structure, with saddle height being a critical factor in determining cyclist-bike fit and joint kinematics [3, 4]. Adjustments in saddle height alter lower-limb joint angles, muscle activation patterns, and ultimately, the distribution of mechanical stress across articular surfaces [2]. For instance, too low saddle height increases the knee flexion, intensifying patellofemoral stress and tension on the quadriceps tendon [3], while an excessively high saddle height increases the range of motion (ROM) of the lower-limb joint angles and tensions the hamstring and calf muscles [4]. Previous study demonstrated that lowering the saddle height, which was measured as the vertical distance from the ischial tuberosity to the medial malleolus, significantly increases tibiofemoral compressive force [5]. On the other hand, the effect of saddle height on joint forces remains controversial. Bini et al. reported greater anterior tibiofemoral forces when cyclists utilized a saddle height greater than their preference [3]. In contrast, another study demonstrated that anterior tibiofemoral forces first increased but then decreased when the saddle height increased from 102% to 120% of the distance between the ischial tuberosity and the medial malleolus [5]. These discrepancies may be attributed to different definitions of saddle height at baseline among the studies. Compounding the above issues, cyclists often seek higher cycling power during their daily training. Although higher intensities and training volumes are essential for performance, inadequate recovery or prolonged exposure to excessive loads may alter lower-limb kinematics (e.g., joint angles, muscle activation patterns) and accelerate fatigue onset [2, 4]. Previous study indicated that the ROM of shank in transverse and the sagittal-plane ROM of foot significantly increased in a higher pedaling intensity [6]. This can increase the valgus and varus loading on the knee and increase the risk for overuse injuries [6]. Furthermore, a study reported that greater pedaling resistance and power outputs significantly increased the compressive and shear forces at the tibiofemoral joint [5]. These findings were substantiated by direct measures of the tibial forces in instrumented knees, in which the results revealed a positive correlation between peak knee forces and power levels [7]. But this invasive experimental approach has significant limitations. Additionally, the interaction between saddle height and power output may play a role in influencing the outcomes but has been greatly underexplored in literature. The relationship between cycling setting, including intensity and structure, and injury is multifaceted. While higher power output is essential for performance, it imposes greater mechanical demands on the musculoskeletal (MSK) system. Prolonged exposure to elevated intensities causes compensatory kinematic adaptations and changed load distribution across joints [8, 9]. Therefore, quantifying muscle and joint forces helps to reveal potential biomechanical connections. However, there are challenges with the quantization method during cycling. Surface electromyography (EMG) has been widely used to infer muscle activation during cycling [10, 11], but it does not correlate well with muscle force due to confounding factors such as changes in fascicle length, contraction velocity, and co-activation patterns. Traditional inverse dynamics for calculating net joint forces are not actual bone-on-bone joint reaction forces [12]. Three-dimensional MSK modeling offers a more robust framework for estimating muscle and joint contact forces [13, 14]. Advancements in model complexity and geometric representativeness have increased the accuracy of force prediction in recent decades [15, 16]. Several studies have successfully developed and validated a model with flexible joints and bike‒human constraints to enable the simulation of cycling [17, 18]. An interaction between the bike structures and muscle forces has been revealed in cycling rehabilitation by the MSK model [19, 20]; for example, the reduced crank length caused greater muscle forces in lower limbs [19]. However, the riding position in a rehabilitative context differs from the normal upper-body position during recreational cycling [20]. Therefore, alternative MSK models tailored for recreational cycling are needed. This study aimed to investigate the influences of saddle height and power output, as well as their interaction effects on lower-limb muscle forces and joint forces, via MSK multibody dynamics simulation. We hypothesized that (1) increased power output would induce greater force outputs of joints and major joint extensors of the lower limb during the propulsive phase, (2) decreased saddle height would result in greater joint forces and muscle forces, and (3) there would be interactive effects between the two factors on the outcome variables. Methodology Participants The inclusion criteria for participants were healthy individuals aged between 20 and 30 years, having a normal body mass index (18.5 ≤ BMI < 24 kg/m²), and being 165–180 cm tall for males or 155–175 cm tall for females. They were required to ride (road bike, mountain bike, or bike trainer) more than once a week for longer than 30 min in daily life. All participants were free from injuries, pain, and MSK disorders for at least six months prior to experiment. Twenty-five healthy participants were recruited. The information about the participants was summarized in Table 1. The greater trochanter height (GTH) was measured as the vertical distance from the greater trochanter to the ground while participants stood in a neutral posture, using a calibrated meter stick. This study was approved by the University Human Subjects Ethics Sub-Committee (Number: HSEARS20220615001). The sample size was estimated in G*power 3.1.9.7 (Heinrich-Heine-Universität Düsseldorf, Düsseldorf, Germany) using multivariate analysis of variance, with a small effect size of 0.25, a β level of 0.8, and an α level of 0.05. Table 1The information about the participantsBasic information Gender15 males, 10 females Age24.64 ± 3.19 years Functional threshold power (FTP)140.09 ± 29.57 wAnthropometric characteristics (mean ± standard deviation) Height170.66 ± 6.99 cm Weight62.10 ± 9.03 kg Body mass index (BMI)21.21 ± 1.82 kg/m2 Greater trochanter height (GTH)87.64 ± 4.91 cmCycling experience (number of participants) Some experience14 Amateur cyclist11Cycling frequency (number of participants) 1 time a week5 2–3 times a week16 4–5 times a week4 ≥ 6 times a week0Average time of every cycling (number of participants) 30–60 min10 1–3 h11 ≥ 3 h4 Table 1The information about the participantsBasic information Gender15 males, 10 females Age24.64 ± 3.19 years Functional threshold power (FTP)140.09 ± 29.57 wAnthropometric characteristics (mean ± standard deviation) Height170.66 ± 6.99 cm Weight62.10 ± 9.03 kg Body mass index (BMI)21.21 ± 1.82 kg/m2 Greater trochanter height (GTH)87.64 ± 4.91 cmCycling experience (number of participants) Some experience14 Amateur cyclist11Cycling frequency (number of participants) 1 time a week5 2–3 times a week16 4–5 times a week4 ≥ 6 times a week0Average time of every cycling (number of participants) 30–60 min10 1–3 h11 ≥ 3 h4 The information about the participants Instruments A motion capture system (T20, Vicon Motion Analysis Inc., Oxford, UK) including eight infrared-based cameras was used to acquire trajectories of markers at a sampling frequency of 250 Hz. Twenty-six reflective markers were placed on the shoulder, torso, pelvis, right and left thighs, shanks, and feet of participant. The mountain bike (Marlin 7 Gen 2, Trek, Intrepid Corporation, USA) was assembled on a cycling training platform (Tacx NEO 2T, Garmin, USA) with the rear wheel removed [21], and the pedals were replaced with a pair of force pedals instrumented with six-component force sensors (I-CrankSet ICS-M, Sensix, Poiters, France). The horizontal position of saddle was fixed at neutral position. The head tube length was 9 cm which fixed the handlebar height. Therefore, the saddle-to-handlebar height difference was determined only by the saddle height. The effective top tube length was 60.2 cm which fixed the saddle-to-handlebar horizontal distance. All other bicycle settings were strictly controlled in the experiment to maintain consistency. A wireless EMG system (Delsys, Boston, MA, USA) was used to collect the EMGs of leg muscles. The markers’ trajectories, pedal forces and torques, and EMGs were recorded synchronously throughout the experiment. Experimental protocol The anthropometric information of the participant was first collected. The maximum power output was tested for four minutes and converted into one hour by the existing formulas to obtain the functional threshold power (FTP) [21]. Instead of using the standard one-hour test, the four-minute test was easier to complete for amateur cyclists and could minimize fatigue accumulation before the formal experiment. Its strong correlation (R2 = 0.95) with the mean power output of the one-hour test indicated its reliability for performance assessment [22]. The experimental power outputs were set according to the individual’s FTP. This ensured that relative effort was matched across participants and avoided overloading specific participants such as less-experience female participant. Following skin preparation, wireless EMG sensors were attached to the belly of rectus femoris (RF), tibialis anterior (TA), biceps femoris (BF), and gastrocnemius (GAS) muscles of the right leg, following SENIAM standard [23]. For each muscle, participants performed 5-second isometric maximal contractions against manual resistance which was designed in SENIAM. A one-minute interval was given between each against action to avoid fatigue. The maximal voluntary contraction (MVC), for standardizing EMG signals, was calculated as the mean of the root-mean-square EMG values over five seconds. Reflective markers were placed after MVC testing. Two additional markers were placed at the center of the outside edge of the left and right pedals, respectively. The participants warmed up for 10–15 min before the cycling tests. The tests included nine riding conditions (3 saddle heights × 3 power output levels). The saddle height, defined as the distance from the highest point of the plane on the saddle to the ground which was set according to the GTH of participants. The 95%, 100%, and 105% of GTH were chosen because height changes of less than 4% of GTH could not significantly affect the joint angles, pedal force, and muscle activity [24]. The saddle height was adjusted by the same experimenter during the experiment using a meter stick according to the calculated saddle height in advance (95%, 100%, and 105% of GTH). The three power outputs were 25%, 50%, and 75% of participants’ FTP.There was enough rest time (at least two minutes) between each of the two-minute riding tests. The cadence was monitored in real time through the Tacx Training app of cycling training platform. The riding state including cadence was projected onto a large screen in front of the participant so that they could adjust and maintain a required cadence of 85–95 revolutions per minute (rpm). The experimental sequence was determined by randomly permuting numbers 1 to 9 using a customized program. Each number represented one specific riding conditions. All cycling tests for each participant were completed in one day. Musculoskeletal multibody model The Twente lower extremity MSK model from a repository of the Anybody Modeling System v.7.4 (AnyBody Technology, Aalborg, Denmark) was modified for cycling analysis. The model consisted of a simplified upper body (head, cervical spine, thorax, and lumbar) and 11 segments representing the lower limbs (pelvis, right and left thigh, patella, shank, foot, and talus). There were 169 Hill-type muscles and 6 joint degrees of freedom (DOFs) of one leg. The DOFs of the lower-limb joints were three rotations in the hip, flexion/extension in the knee, plantar/dorsal flexion in the ankle, and eversion/inversion in the subtalar joint. No additional DOF was introduced for the patella because its position was defined as a function of the knee flexion angle. Two virtual muscles were added symmetrically to the left and right feet to simulate the support of pedals. The initial and final 30 s of the 2-min cycling data were excluded. Five pedaling cycles between 30 s and 90 s of 2-min riding were selected for each condition. A pedaling cycle was defined as the period from the top dead center (TDC) to the next TDC, where the propulsive phase was 0–180° and the recovery phase was 180–360°. The position of the right foot was identified via the coordinates of the markers attached to the middle axis of the right pedal. The measured pedal reaction forces and torques, as well as the trajectories of the markers obtained from the motion capture system, were input for the dynamic simulation. In accordance with the length–mass–fat scaling law, anthropometric data were used to scale and adjust the segment dimensions, the principal axes, and the moments of inertia for each participant. The weighted least squares method was applied to minimize tracking errors in marker trajectories. The simulation consisted of inverse kinematic and inverse dynamics. The kinematic parameters calculated in the first step were used in dynamic analysis to obtain the muscle forces and joint forces. The sum of muscle activations cubed was minimized based on the least fatigue criterion in the physiological muscle recruitment pattern [25]. The MSK multibody model was verified by comparing the predicted muscle activations of RF, TA, BF, and GAS with the measured EMGs. Figure 1 illustrates the experimental procedure and MSK multibody simulation workflow. Fig. 1The workflow of the cycling experiment and musculoskeletal multibody simulation. (FTP, functional threshold power; MVC, maximum voluntary contraction; EMG, electromyography) The workflow of the cycling experiment and musculoskeletal multibody simulation. (FTP, functional threshold power; MVC, maximum voluntary contraction; EMG, electromyography) Outcome variables The raw EMG signals were filtered using a 4th-order bandpass Butterworth filter with cutoff frequencies of 10–500 Hz. Then, it was full-wave rectified and filtered using a moving root-mean-square with a moving window of 70 ms. The processed EMG was normalized by the MVC values. The EMG envelope was averaged across five cycles and compared to the muscle activation obtained from MSK simulation. The sagittal-plane joint angles of the hip, knee, and ankle, as illustrated in Fig. 1, were calculated and exported by Vicon Nexus 2.16 using the lower body model. The joint angles for each pedaling cycle were normalized to the crank angle (0-360°) as the horizontal coordinate variable. For the specific joint of each participant, the joint angle curve represented the average angles of all pedaling cycles under a specific riding condition. From the curve, the maximum, minimum, and ROM of joint angles were determined. The muscle forces of RF, TA, BF, GAS, soleus (SOL), vastus lateralis (VL), vastus medialis (VM), and gluteus maximus (GluMax) were selected from the calculated results. The component joint forces of the hip, knee, and ankle in the medial and lateral (M-L), proximal‒distal (P-D), and anterior-posterior (A-P) directions were output for analysis. All the variables were expressed as functions of crank angle with degrees (0-360°) and were averaged across five pedaling cycles and the study cohort. Statistical analysis The MSK model was validated by examining the agreement between the processed EMG data and calculated muscle activations via the Pearson correlation coefficient r. The agreement was classified as strong (|r| ≥ 0.7), moderate (0.4 ≤ |r| < 0.7), low (0.1 ≤ |r| < 0.4), or negligible (|r| < 0.1) [26]. Normality of the kinematic data was checked using Shapiro-Wilk tests (p > 0.05). Sphericity was examined via Mauchly’s test with Greenhouse-Geisser corrections (p > 0.05). A two-way repeated measures analysis of variance (ANOVA) was conducted to explore the effects of saddle height and power output on hip, knee, and ankle joint angles. Significant interactions were followed up with simple effects analyses, and pairwise comparisons were adjusted using Bonferroni correction. The mean differences and the 95% confidence intervals (CIs) were calculated, along with the statistical significance level (p < 0.05). The Shapiro‒Wilk test revealed that the data of muscle forces and joint forces were not normally distributed (p < 0.05). A generalized estimating equation (GEE) model with a gamma distribution and a log link was used to explore the associations between the experimental variables and outcomes. The degrees of associations were estimated by the Wald statistic for a power output × saddle height interaction term, with separate estimations if ɑ < 0.05 for the interaction term. An exchangeable covariance structure was employed. Each model was adjusted by gender, BMI, and cadence. The null models were 25% FTP for power output and 95% GTH for saddle height. The model performance was assessed by Quasilikelihood under the independence model criterion. The direction and magnitude of association were assessed by effect β and risk ratios with 95% CIs, which quantify the relative change in outcomes per unit predictor (5% GTH of saddle height and 25% FTP of power output). The significance was set at p < 0.05 with Holm‒Bonferroni correction for post hoc comparisons. All the statistical analyses were performed with IBM SPSS statistics 26.0 (IBM Corporation, Armonk, NY, USA). Results Validation of the musculoskeletal model Analyses of the Pearson correlation coefficient demonstrated a strong correlation between the calculated and measured muscle activations in the TA (r = 0.960, p < 0.001) and BF (r = 0.967, p < 0.001), as shown in Fig. 2. A moderate correlation was found between RF (r = 0.513, p < 0.001) and GAS (r = 0.496, p < 0.001). Table 2 summarizes the comparisons of peak value, peak timing (represented as the crank angle), peak cycling phase, and duration of muscle activation (expressed as the proportion of a pedaling cycle) between the calculated and measured muscle activations. The pedaling cycle was divided into four phases based on the crank angle: phase 1 (0° − 90°), phase 2 (90° − 180°), phase 3 (180° − 270°), and phase 4 (270° − 360°). The calculated activation peaks of all the muscles were generally lower than those of the EMG; for example, the calculated peak activation of the RF muscle was 6.74% greater than its measured EMG. The difference in the peak timing between the calculations and measurements of the muscle activations was less than 10°, except for that of RF. Besides, RF also had the largest difference between calculated and measured activation duration, reaching 9.17%, whereas the differences in the other three muscles were less than 5%. Fig. 2The processed electromyography (EMG) and model-calculated muscle activation of the rectus femoris (RF), tibialis anterior (TA), biceps femoris (BF), and gastrocnemius (GAS) muscles The processed electromyography (EMG) and model-calculated muscle activation of the rectus femoris (RF), tibialis anterior (TA), biceps femoris (BF), and gastrocnemius (GAS) muscles Table 2Comparison between processed EMG data and calculated muscle activationsPeak value (% MVC)Peak timing (°)Cycling phaseMuscle activation (%)Correlation coefficientRFEMG23.5794.20261.390.513Model16.8326.53152.22TAEMG19.16349.47445.560.960Model13.55341.60443.33BFEMG36.1095.73248.340.967Model20.6688.10153.32GASEMG29.61165.47253.060.496Model18.13156.40256.94RF rectus femoris, TA tibialis anterior, BF biceps femoris, GAS gastrocnemius, EMG electromyography, MVC maximum voluntary contraction Table 2Comparison between processed EMG data and calculated muscle activationsPeak value (% MVC)Peak timing (°)Cycling phaseMuscle activation (%)Correlation coefficientRFEMG23.5794.20261.390.513Model16.8326.53152.22TAEMG19.16349.47445.560.960Model13.55341.60443.33BFEMG36.1095.73248.340.967Model20.6688.10153.32GASEMG29.61165.47253.060.496Model18.13156.40256.94RF rectus femoris, TA tibialis anterior, BF biceps femoris, GAS gastrocnemius, EMG electromyography, MVC maximum voluntary contraction Comparison between processed EMG data and calculated muscle activations RF rectus femoris, TA tibialis anterior, BF biceps femoris, GAS gastrocnemius, EMG electromyography, MVC maximum voluntary contraction Joint angles The hip, knee, and ankle joint angles and ROMs were significantly differences among the three saddle heights (Fig. 3). The specific statistical analysis results were detailed in Supplementary Table 1. The maximum (flexion) and minimum (extension) of hip joint angle were decreased by 3.58° (p = 0.008) and 3.89° (p = 0.001), respectively, as the saddle height increased from 95% to 105% of GTH. As for knee flexion angle, the maximum angle at 95% GTH reduced by 7.52° (p < 0.001) and 13.12° (p < 0.001) compared to that at 100% GTH and 105% GTH. The minimum knee joint angle decreased more to reach 11.48° (p < 0.001) and 24.81° (p < 0.001), accordingly. The ROM of knee angle was most affected by saddle height among lower-limb joints. It increased by 3.96° (p = 0.012) and 11.69° (p < 0.001), respectively, when the saddle height was raised to 100% and 105% of GTH from 95% GTH. Regarding the ankle joint, the minimum joint angle (dorsiflexion) decreased as the saddle height increased (p < 0.05), which were 4.32° at 95% GTH, -0.10° at 100% GTH, and − 6.99° at 105% GTH. No significant change was found in its maximum. The raised saddle height increased the ROM of the ankle joint angle by 4.67° (100% GTH compared to 95% GTH, p = 0.027) and 8.58° (105% GTH compared to 95% GTH, p = 0.005). Fig. 3The hip, knee, and ankle joint angles at three saddle heights (95%, 100%, and 105% GTH) with power output of 50% FTP. (SD, standard deviation) The hip, knee, and ankle joint angles at three saddle heights (95%, 100%, and 105% GTH) with power output of 50% FTP. (SD, standard deviation) At a specific saddle height, the power output has less influence on the hip and ankle joint angles than saddle height and did not show a consistent variation pattern. The knee joint angle was significantly changed (p < 0.05) by power output only at a saddle height of 100% GTH. The minimum knee joint decreased by 1.43° (50% FTP compared to 25% FTP, p < 0.001) and 3.02° (75% FTP compared to 25% FTP, p < 0.001) as the power output increased. The ROM of knee angle increased by 2.01° (p = 0.039) as the power output increased from 25% to 75% of FTP, but there was no significant change at 50% FTP. Muscle forces Figure 4 displays the mean muscle forces at different power outputs. As the power output increased from 25% to 50% of FTP, the muscle forces of the VL (β = 0.014, p = 0.022), VM (β = 0.013, p = 0.026), BF (β = 0.053, p = 0.005), and GluMax (β = 0.025, p < 0.001) increased significantly, whereas those of the GAS (β = -0.142, p < 0.001) and SOL (β = -0.060, p = 0.001) decreased. When the output power increased to 75% FTP, the muscle forces of the VL, VM, BF, and GluMax increased by 8.03 N (β = 0.018, p = 0.010), 3.59 N (β = 0.016, p = 0.017), 9.44 N (β = 0.119, p < 0.001), and 8.56 N (β = 0.042, p < 0.001), respectively. TA also showed an increase from 17.24 N to 19.81 N (β = 0.208, p = 0.010). The mean muscle forces of the GAS (β = -0.268, p < 0.001) and SOL (β = -0.107, p = 0.001) decreased by 11.0 N and 13.67 N, accordingly. Fig. 4Mean muscle forces of the rectus femoris, vastus lateralis, vastus medialis, biceps femoris, gastrocnemius, soleus, tibialis anterior, and gluteus maximus, with cycling power outputs of 25%, 50%, and 75% of the functional threshold power (FTP). * represents statistical difference between the results of two riding conditions (p < 0.05). o represents outlier data. The line in the box indicates the middle value of the dataset. × in the box indicates the mean value Mean muscle forces of the rectus femoris, vastus lateralis, vastus medialis, biceps femoris, gastrocnemius, soleus, tibialis anterior, and gluteus maximus, with cycling power outputs of 25%, 50%, and 75% of the functional threshold power (FTP). * represents statistical difference between the results of two riding conditions (p < 0.05). o represents outlier data. The line in the box indicates the middle value of the dataset. × in the box indicates the mean value Figure 5 illustrates the effects of saddle height on the mean muscle forces. The forces of RF, VL, VM, and GluMax significantly decreased with each 5% GTH increase in saddle height. When the saddle height increased from 95% to 105% of GTH, the muscle forces of the RF (β =-0.096, p < 0.001), VL (β =-0.032, p = 0.007), VM (β =-0.036, p = 0.003), GAS (β =-0.214, p = 0.026), and GluMax (β =-0.050, p < 0.001) decreased, whereas that of the TA increased (β = 0.208, p = 0.010) by 32.96% from 15.9 N to 21.2 N. The detailed GEE analysis results of the effects of power output and saddle height on the muscle forces are included in Supplementary Table 2. Fig. 5Mean muscle forces of the rectus femoris, vastus lateralis, vastus medialis, biceps femoris, gastrocnemius, soleus, tibialis anterior, and gluteus maximus with saddle heights of 95%, 100%, and 105% greater trochanter height (GTH). * represents statistical difference between the results of two riding conditions (p < 0.05). o represents outlier data. The line in the box indicates the middle value of the dataset. × in the box indicates the mean value Mean muscle forces of the rectus femoris, vastus lateralis, vastus medialis, biceps femoris, gastrocnemius, soleus, tibialis anterior, and gluteus maximus with saddle heights of 95%, 100%, and 105% greater trochanter height (GTH). * represents statistical difference between the results of two riding conditions (p < 0.05). o represents outlier data. The line in the box indicates the middle value of the dataset. × in the box indicates the mean value The GEE model suggested that cadence was a significant confounder for the BF (β = 0.012, p = 0.004), GAS (β = 0.020, p = 0.014), and GluMax (β = 0.004, p < 0.001) muscle forces. A greater cadence tended to result in greater muscle forces. BMI was significantly associated with the BF (β = 0.112, p < 0.001), TA (β = 0.214, p = 0.004), and GluMax (β = 0.043, p < 0.001) muscle forces. Gender affected only the BF muscle force (β = -0.287, p = 0.008). Joint forces The maximum joint forces with various power outputs are compared in Fig. 6. When the power output increased from 25% to 75% of FTP, the three components of the hip joint force increased significantly from 297.7 N to 313.2 N (M-L component), from 1327.4 N to 1436.7 N (P-D component), and from 119.2 N to 131.2 N (A-P component), resulting in a 9.37% increase in its resultant force (β = 0.074, p < 0.001). The P-D and A-P ankle joint forces were reduced by 32.43 N (β = -0.048, p < 0.001) and 9.64 N (β = -0.047, p = 0.037), respectively, as the power output increased from 25% to 50% FTP.The P-D ankle joint force further decreased from 644.1 N to 593.2 N (β = -0.080, p < 0.001) as the power output increased to 75% FTP.GEE analysis revealed an interaction effect (p = 0.030) of saddle height and power output on the M‒L knee joint force. The influence of power output on the M-L knee joint force was compared at specific saddle height. The three groups from left to right in Fig. 6 are 95%, 100%, and 105% of GTH. The maximum of M-L knee joint force showed a slight increase with increasing power output at three saddle heights (β = 0.002–0.028), but there was no statistical difference. The knee joint forces in P-D and A-P directions also did not change significantly. Fig. 6Maximum joint forces of the hip, knee, and ankle along the medial and lateral (M-L), proximal‒distal (P-D), and anterior-posterior (A-P) directions with cycling power outputs of 25%, 50%, and 75% of the functional threshold power (FTP). * represents statistical difference between the results of two riding conditions (p < 0.05). o represents outlier data. The line in the box indicates the middle value of the dataset. × in the box indicates the mean value Maximum joint forces of the hip, knee, and ankle along the medial and lateral (M-L), proximal‒distal (P-D), and anterior-posterior (A-P) directions with cycling power outputs of 25%, 50%, and 75% of the functional threshold power (FTP). * represents statistical difference between the results of two riding conditions (p < 0.05). o represents outlier data. The line in the box indicates the middle value of the dataset. × in the box indicates the mean value The effects of saddle height on the maximum joint forces are displayed in Fig. 7. The maximum hip joint force in the A‒P direction decreased as the saddle height increased from 95% to 100% (β = -0.064, p = 0.020) and further to 105% (β = -0.125, p < 0.001) of GTH. The A‒P knee joint force decreased only when the saddle height improved by 10% GTH (β = -0.070, p = 0.008). Owing to this interaction, the effects of saddle height on M‒L knee joint forces were examined at each power output level. When the power output was 75% FTP, an increase of 10% GTH in the saddle height resulted in a decrease of 13.13 N in the maximum M‒L knee joint force. The maximum of the resultant knee joint force was only slightly reduced (β = -0.009, p = 0.546). The maximum ankle joint forces in the M-L (β = -0.126, p = 0.045) and A-P (β = -0.089, p = 0.001) directions were lower at a saddle height of 105% GTH than that at 95% GTH. Supplementary Table 3 displays the specific GEE results of the influence of power output and saddle height on the joint forces. Fig. 7Maximum joint forces of the hip, knee, and ankle along the medial and lateral (M-L), proximal‒distal (P-D), and anterior-posterior (A-P) directions with saddle heights of 95%, 100%, and 105% greater trochanter height (GTH). * represents statistical difference between the results of two riding conditions (p < 0.05). o represents outlier data. The line in the box indicates the middle value of the dataset. × in the box indicates the mean value Maximum joint forces of the hip, knee, and ankle along the medial and lateral (M-L), proximal‒distal (P-D), and anterior-posterior (A-P) directions with saddle heights of 95%, 100%, and 105% greater trochanter height (GTH). * represents statistical difference between the results of two riding conditions (p < 0.05). o represents outlier data. The line in the box indicates the middle value of the dataset. × in the box indicates the mean value The GEE model revealed that BMI and cadence were significant confounders of joint forces. Individuals with a higher BMI appeared to have greater joint forces, especially for the hip joint. A faster cadence also tended to cause greater joint forces, especially the maximum ankle joint force. Discussion The mean muscle force in a pedalling cycle can better reflect the level of muscle activation than the peak value. The maximum joint force is strongly associated with injury risk [24, 27]. The results partially supported our hypotheses. Specifically, as the power output increased, most muscle forces (mean values), hip joint forces, and knee joint forces (maximum values) increased. In contrast, the ankle joint forces and calf muscle engagement declined. With increasing saddle height, we observed a general reduction in the lower-limb muscles except for the TA. The mean forces exerted on all the joints decreased, and there was a notable reduction in the maximum forces at the hip and ankle joints in the A‒P direction. The interaction effect between power output and saddle height significantly affected only the M‒L knee joint force and the GluMax muscle force. Musculoskeletal model for cycling A multibody MSK model for cycling was compiled. The moderate to strong agreement between the calculated results and EMG measurements of the selected muscles demonstrates the validity of the model. Among them, the lowest correlation coefficient for the GAS may result from the difference between the measured and calculated target muscles. The calculated muscle activation of the GAS included the medial head and the lateral head, whereas the EMG sensor was attached to the GAS medial. Despite this, our model incorporates more lower-limb muscles than previous cycling models [19, 28]. The phase delay between the activation timing and model prediction occurred in the selected muscles, which could be expected because of the physiological electromechanical delay phenomenon. Muscle forces influenced by power output In accordance with previous studies [29–31], greater power output resulted in greater muscle forces across the major lower-limb muscles. Since hip and knee extensors are the major contributors during the propulsive stage, it is unsurprising that both muscle groups elicited greater forces as the power output increased. This corresponds to the increase in the ROMs of hip and knee angles. Regardless, the increments of muscle forces were not evenly distributed among synergist muscles. For instance, the VL and VM displayed increases in their mean muscle forces, whereas the RF showed an increase in only the maximum value. The RF, as a biarticular muscle, mainly facilitates force transmission across joints [32]. VL and VM, as monoarticular muscles, are responsible for power generation to meet the demands of increased exercise intensity [33]. Conversely, the SOL and GAS exhibited less force production as the power output increased, which was inconsistent with previous findings, as some studies reported increased EMG [34], whereas others reported no significant change in activation in the same muscle group [30]. These discrepancies are likely attributed to differences in bike construction, such as the position of cleats and power output setting methods. Our study utilized FTP to gauge power output, a metric that emphasizes sustained force production, which is generally lower than the maximum power of the incremental cycling test used in the previous study. In fact, the SOL consists predominantly of slow-twitch fibers and limited fast-twitch fibers [35]. Cyclists tend to shift more toward muscles containing more fast-twitch fibers under greater power output, such as the quadriceps and gluteal muscles [36]. Therefore, the muscle forces of the RF and BF increased, whereas that of the SOL decreased. Muscle forces influenced by saddle height One of the most notable changing trends was the decrease in the mean forces of the VL, VM, and RF. They are synergistic knee extensors and are highly engaged in the propulsive phase. On the basis of the muscle length‒force relationship principle, the knee extensors are at the optimal length for maximum force production when the knee is flexed at 60–70° [37]. The saddle height of 100%–105% GTH shifted the knee flexion angle into this range. The knee extensors could enter the optimal working region and therefore reduce their activation intensities. Owing to the kinematic chain, increased saddle height is coupled with reduced knee flexion and increased moment arms of the knee extensors to the knee joint, resulting in less force demand on the knee extensors for producing the same joint torque during propulsion. Similar mechanisms can explain the reduced mean forces in other propulsive muscles (GlueMax, BF, SOL, and GAS). The hip joint exhibited decreased flexion and extension as saddle height increased, likely altering the working lengths of the hip extensors (e.g., GlueMax) and flexors (e.g., RF). These results are in line with previous studies that demonstrated a negative correlation between saddle height and lower-limb muscle activations [38, 39]. In contrast, the TA force increased as the saddle height increased which aligns with a study that adjusted the saddle height on the basis of the knee flexion angle [40]. Raising the saddle height led to a more dorsiflexed position of ankle. Since TA and GAS act as antagonistic muscles, the increase in TA muscle force corresponds to the reduction in GAS force. Joint forces influenced by power output The resultant joint forces of the hip and knee increased with increasing power output, whereas those of the ankle decreased, which was mainly due to the reductions in the P-D and A-P component forces. Previous studies documented increased power and joint loadings of the hip and knee joints at higher power outputs [5, 7, 41]. This increase can be attributed to the increased muscle forces across the two joints [42]. Conversely, our findings regarding the ankle joint force differed from research reporting that adding 120 W to the power output significantly increased the ankle compressive force [43]. The study assumed that the ankle compressive force was the sum of the component pedal force along the long axis of the tibia and the Achilles tendon force. This assumption compromised the strength of the outcomes because of the imprecise estimations of the moment arm and oversimplified computations without considering muscle forces. In our study, the reduction in SOL and GAS muscle forces likely contributed to the decreased ankle loading. Further evidence to support this notion is that, under the condition of greater physical demands, the contributions of the knee and hip joints were found to increase substantially, whereas those of the ankle joint remained unchanged or even decreased [44]. Joint forces influenced by saddle height The reduction in the maximum A‒P joint forces at the hip, knee, and ankle can be attributed to the changes in joint kinematics induced by a higher saddle height. As the saddle height increased, the hip joint exhibited decreased flexion and extension, which likely contributed to a more upright pedaling posture. This reduced hip angulation aligns with previous findings that higher saddle positions reduce the forces resolved onto the shear planes [24]. Notably, the results indicated a reduction in knee force of approximately 11.7% per 5% GTH increase in the saddle height, closely matching a study of in-vivo hip joint loading measured by implants which reported 7%–15% decreases per 9 cm increase in saddle height [42]. This good comparability supports the biomechanical consistency across studies, particularly given the similar range of saddle height adjustment in this study (about 8.76 cm). At the knee joint, the reduction in maximum flexion and extension angles led to a greater ROM. Similarly, this more extended knee position results in declined forces in the tangent plane. On the other hand, the reduced muscle forces of the quadriceps and hamstrings also contributed to the decreased hip and knee joint forces [45]. Our results showed that a greater saddle height may be beneficial to cyclists. Both muscle forces and joint forces increase when the saddle angle is low since it results in high flexion angles at the hip, knee, and ankle joints [46]. However, excessive saddle height may lead to perineal numbness and Achilles tendinopathy [47]. Therefore, there must be a preferable range for setting the highest feasible saddle for the sake of minimizing injuries. Scholars have proposed that the appropriate saddle setting would be knee flexion of 25° to 30° at the bottom dead center [47]. Nevertheless, the criteria to anchor the saddle height should consider additional factors, such as the bike design and riders’ competitive levels. Our study also demonstrated that power output and cadence influenced loading on the lower limb. Therefore, the optimal saddle height is likely a decision on an individual basis that balances injury prevention and performance demands. Limitations This study has several limitations. First, since the sample population consisted solely of young amateur cyclists, the outcomes may not be generalizable to cyclists of other ages or professional athletes. Elite cyclists with high training volumes were found to adopt different muscle recruitment strategies compared to amateur cyclists [48]. Second, we used mountain bikes in the experiment to enhance the cycling stability and the comfort of participants. Therefore, the results could not be compared with studies using other bicycle types. Third, we employed a smaller cycling power output than other studies did [29, 49] to minimize the confounding effects of muscular fatigue. This approach does not account for the full range of strenuous activities in real-world riding conditions and may limit the significance of the observed differences between conditions. Last, the calculated activation of the RF and GAS showed only moderate agreement with their EMG data. This may be related to the added virtual muscles in the shank, which could bear part of the loads. In the future, comparing other bicycle types and exploring more settings on the bicycle structure would be necessary to define the optimal adjustment. The moment of peak muscle forces and joint forces should be investigated, as it is significant for further understanding the influence of saddle height and power output on the coordination pattern of lower limbs. Conclusion In conclusion, this study used a validated MSK multibody model to quantify how saddle height and power output influence lower-limb muscle and joint forces during stationary cycling. The calculated results of kinematics, including the joint angles and ROM of hip, knee, and ankle, revealed the reason for the changes of kinetics. The increased power output significantly increased the mean forces in most muscles, particularly in the hip and knee extensors, whereas the forces in the ankle joint and calf muscles decreased. Additionally, increasing the saddle height reduced most muscle forces, except for the TA. This is one of the reasons for the reductions in hip, knee, and ankle joint forces as the saddle height increases. However, these findings must be interpreted with caution: while higher saddle heights may reduce lower-limb MSK loading, optimal adjustments should also consider other factors not evaluated in this study, such as saddle pressure distribution. Future research on these variables is necessary to improve the guidance for saddle height adjustment. Supplementary Information Supplementary Material 1. Supplementary Material 1. Supplementary Material 2. Supplementary Material 2. Supplementary Material 3. Supplementary Material 3.