Redefining muscular action: human “adductor” magnus is designed to act primarily for hip “extension” rather than adduction in living young individuals
Abstract
Human leg muscles are uniquely enlarged for upright bipedalism, and the adductor magnus is one of the largest muscles. Although this muscle is recognized as a hip adductor, hip adduction torque is not greatly required during human locomotion, such as walking and running. The functional role of this giant muscle remains unclear. Here, we tested the hypothesis that the human adductor magnus acts primarily for hip extension rather than adduction in living young individuals. Utilizing diffusion tensor imaging, we reconstructed fascicles over the entire muscle in 15 young adults at the hip neutral position. We divided the muscle into three portions based on fascicle insertion and examined their three-dimensional architectures. The posterior and anterior-distal portions comprised over 80% of the whole muscle volume and physiological cross-sectional area. These portions demonstrated a longer moment arm for hip extension than adduction. Consequently, the potential torque (maximal torque-generating capacity) of the whole muscle was over twofold greater for hip extension than adduction. The hip extension potential torque was correlated with the maximal hip extension torque measured with a dynamometer. These results highlight the architectural design of the adductor magnus, favoring hip extension over adduction, providing novel insights into its functional role beyond the frontal plane in human locomotor mechanics.
NEW & NOTEWORTHY The human adductor magnus, one of the largest leg muscles, is traditionally considered a hip “adductor.” However, its functional role is unclear. We found that the torque-generating capacity of this muscle for hip extension was substantially greater than that for hip adduction and explained the exerted torque for hip extension. Our findings highlight the important role of the adductor magnus as a major hip “extensor,” having implications for mechanisms of human locomotion and musculoskeletal simulations.
INTRODUCTION
Human lower limb muscles are uniquely enlarged as compared with those of apes, potentially reflecting our evolutionary adaptation to terrestrial bipedal locomotion (1–4). The adductor magnus (AM) is one of the largest muscles in the human lower limb (5, 6). Traditionally, the AM is known as a major hip “adductor,” playing a pivotal role as a frontal-plane stabilizer of the pelvis in human movements, along with other hip adductors (7). However, hip abduction torque greater than the hip adduction torque is required for stabilizing our pelvis in the frontal plane during a single-leg stance in human locomotion, including walking and running (8–10), although a certain amount of adduction torque may be needed to increase joint stability. Accordingly, there can be a small demand for hip adduction torque generation in human locomotion, for which humans are known to have enlarged muscles during evolution (2). This raises fundamental questions: Why is the human AM so large? Is it primarily an adductor?
Cadaveric studies on human AM (11–13) have identified at least two functional compartments within the muscle: the pubofemoral (anterior) and ischiocondylar (posterior) portions. The anterior portion (further divided into proximal and distal portions), accounting for approximately 70% of the total AM volume and physiological cross-sectional area [PCSA; (11)], originates from the pubis and inserts onto the femur, primarily acting as a hip adductor according to its moment arm (14). In contrast, the posterior portion arises from the ischial tuberosity and extends to the medial femoral epicondyle, contributing to hip extension (14). However, these findings are based on a limited number of elderly cadavers who potentially exhibit decreased physical activities in daily life and altered muscle architectures from young individuals (15), although they are relevant and widely used for musculoskeletal simulation models. It remains unclear whether the anatomical findings obtained from elderly cadavers can be applied to living young individuals who are more physically active.
An intramuscular electromyographic study on living young individuals reported that posterior and anterior portions of AM were more active during hip extension than adduction (16). This is supported by another study using surface electromyography (17). Besides, human locomotion requires much greater hip extension torque than adduction torque (10, 18). It is thus hypothesized that AM is designed to act primarily for hip extension rather than adduction in living young individuals, as discussed recently (19). Nevertheless, less information is available regarding the AM function from the perspective of the in vivo architecture.
Numerous studies have quantified human muscle architecture in vivo using B-mode ultrasonography (20–23). However, due to the limited field of view of ultrasound imaging, previous research has primarily focused on certain muscle groups, such as the triceps surae (24) and quadriceps femoris (25). The architecture and potential actions of other large muscles, including AM, remain poorly understood in living humans. In this study, we comprehensively examined the three-dimensional (3-D) architecture and potential actions of human AM in living young individuals through a novel approach based on diffusion tensor imaging (DTI) and tractography, which enables 3-D visualization and quantification of muscle fascicles over a large region of interest in vivo (26, 27).
METHODS
Participants
An a priori power analysis was performed on G*Power software to determine the sample size required for detecting significant differences in muscle architectural parameters among portions and hip joint actions. The effect size (Cohen’s d) was set to be 0.8 (large) with an α level of 0.05 and a power of 0.80. As a result, the required sample size was estimated to be 15. Thus, we recruited a total of 15 healthy adults (10 males, age: 27.5 ± 3.8 yr, body mass: 68.1 ± 9.5 kg, body height: 170.1 ± 5.2 cm; and 5 females, age: 23.4 ± 1.1 yr, body mass: 58.6 ± 9.2 kg, body height: 159.9 ± 5.7 cm, means ± standard deviation). The participants were sedentary or physically active, and they did not have a history of AM injury. The purpose, burdens, and potential risks of this study were explained to the participants, and we obtained written informed consent from them. This study was approved by the Institution’s Research Ethics Committee (Reference Number: 23004).
Image Acquisition
Participants lay prone with their hips and knees fully extended inside the magnet bore of a 3-T MR scanner (MAGNETOM Skyra, Siemens Healthineers, Germany). Their right thigh was positioned without any internal or external rotation, adduction, or abduction of the hip joint. Participants remained in the prone position for at least 20 min before the first MR scan to reduce the fluid shift effect on lower limb muscle morphology due to postural changes (28). To control for potential effects of contraction and/or stretching history on muscle slackness, specifically the fascicle architecture in a given position (29), participants performed a 2–3-s isometric hip extension at a subjective contraction intensity of 50% before the first scan. Then, they were instructed not to change the lower limb position until the end of all scans, and their legs were secured by wrapping the distal shank with a nonelastic band if needed.
A series of axial MR images of the hips and thighs were acquired using a 32-channel spine array coil and two 18-channel body array coils. Two sequences covered the same field of view from the hips to the thighs: T1 VIBE two-point Dixon [gradient recalled echo, echo times (TE): 2.46/3.69 ms, repetition time (TR): 5.98 ms, slice interval: 5 mm, field of view: 315 × 420 mm, reconstructed matrix size: 384 × 512, bandwidth: 610 Hz, flip angle: 12°, and scan time: 19 s for a single scan] and DTI (spin echo-echo planar imaging, TE: 60 ms, TR: 8,300 ms, slice interval: 5 mm, field of view: 315 × 420 mm, reconstructed matrix size: 180 × 240, number of averages: 2, b = 0 and 500 s/mm2, 12 gradient directions on a hemisphere, bandwidth: 3,206 Hz, flip angle: 90°, and scan time: 408 s for a single scan). We repeated both Dixon and DTI scans three times to cover the whole hips and thighs. Participants were instructed to remain fully relaxed during the scanning and to hold their breath during the Dixon scan for the hips to minimize respiration-induced motion artifacts.
Muscle Architecture Analysis
The workflow of the muscle architecture analysis is shown in Fig. 1 and is similar to a previous study (30). In the Dixon (in-phase and water) images, the AM muscle belly as a whole and the femoral head ridges in the right leg were manually segmented by a single examiner (K.T.) using a 3-D slicer (31). In addition, several voxels were segmented around the adductor hiatus (the point where AM splits into two portions) and the anterior/posterior superior iliac spines (ASISs/PSISs, the most anterior/posterior point of the proximal half) of both sides of the pelvis and the medial/lateral epicondyles (ME/LE, the most medial/lateral point of the distal half) of the right femur. These segmentations were resized to a matrix size of diffusion-weighted images in the DTI sequence and imported to a DSI studio (32) in subsequent analyses.
Figure 1.The workflow of muscle architecture analysis. AM, adductor magnus; DTI, diffusion tensor imaging; DW, diffusion weighted; PCSA, physiological cross-sectional area.
The diffusion-weighted images obtained in the DTI sequence were denoised using local principal component analysis (33). The denoised images were then imported into DSI studio (32) for calculation of the primary eigenvector and fractional anisotropy (FA) for each voxel. Fiber tracking was performed using the primary eigenvectors via a deterministic tractography algorithm (32). The tracking began bidirectionally at a 10-mm step from a seed point randomly placed within the whole region of the AM muscle belly segmentation. The fiber tracking was terminated if any of the following criteria were met: FA < 0.1, the angle between two consecutive steps (turning angle) > 50°, or if the fiber extended outside the muscle belly. In addition, fibers shorter than 50 mm or longer than 500 mm were excluded. These parameters were chosen in preliminary studies to maximize the visibility of fibers within the muscle while minimizing the number of unrealistic fibers, referring to previous anatomical studies (11–13). Furthermore, some studies have reported that DTI-based muscle architecture measurements are less sensitive to fiber-tracking parameters (34–35). Fiber tracking was repeated until 5,000 fibers were extracted for a single muscle.
The 3-D coordinate data of all the segmentations and tracked fibers were imported into MATLAB R2023a (MathWorks, Natick). The tracked fibers were trimmed to fall within the muscle belly, and these trimmed fibers were assumed as fascicles. It should be noted that previous DTI studies confirmed the validity that tracked fiber architecture represents muscle fascicle architecture (26, 36, 37). Cadaveric studies on the detailed architecture of the human AM divided this muscle into four portions (“AM1,” “AM2,” “AM3,” and “AM4” in their study) according to positions of the adductor hiatus and perforating arteries between portions (11). However, we can identify only the adductor hiatus in corresponding in-phase images. Thus, we first divided this muscle into the posterior portion (AM4) with fascicle insertions medial to the adductor hiatus (geometric center of all constituent voxels) and the anterior portion (AM1, AM2, and AM3) with fascicle insertions lateral to the adductor hiatus. Meanwhile, the other cadaveric study reported that the AM1 was located above one-fourth of the distance between the lesser trochanter and adductor tubercle (38). This distance is almost equivalent to the distance between the most proximal and distal position coordinates of the muscle belly (muscle belly length). Therefore, we further divided the anterior portion into the “anterior-proximal” (AM1) and “anterior-distal” (AM2 and AM3) portions. The anterior-proximal fascicles were defined as those whose insertions were located between the most proximal point of the muscle belly and one-fourth of the total muscle belly length. The remaining fascicles were defined as anterior-distal fascicles with insertions lateral to the adductor hiatus. The example of fascicles in each portion is shown in Fig. 2.
Figure 2.Typical examples of 3-D reconstruction of muscle fascicles in the human adductor magnus. 3-D, three-dimensional; Ant-dist, anterior-distal portion; Ant-prox, anterior-proximal portion; Post, posterior portion.
Fascicle length was calculated as the sum of distances between consecutive points along the fascicle. The median length within each portion was used for further analyses. Consistent with prior studies (39, 40), the hip joint center of rotation (CoR) was determined by performing a least-squares sphere fitting on the 3-D position coordinate of the femoral head ridges. It should be noted that this is a static method and thus does not account for potential changes in joint CoR during movements. Using the 3-D position coordinates of anatomical landmarks (geometric centers of all voxels of ASISs, PSISs, ME, and LE) and the hip joint CoR, the coordinate systems for the pelvis, femur, and hip joint were defined according to the methodology described in a previous study (41).
For each fascicle, the line of action was approximated as a single unit vector passing through the fascicle origin and midpoint along its length. This fascicle line of action vector was projected onto the plane orthogonal to each rotational axis of the hip joint (Fig. A in Supplemental Supporting Information S1). The length of the projected line of action vector was then calculated as the force fraction, which represents the proportion of fascicle force contributing to joint torque in each plane, ranging from 0 to 1. The moment arm length about each rotational axis was calculated as the shortest distance between the hip joint CoR and line of action vector projected in the orthogonal plane, similar to the method for calculating the tendon moment arm in a previous study (42). The specific torque (an estimate of torque-generating capacity for a given cross-sectional area) of each fascicle was calculated by multiplying the product of the force fraction and moment arm length by the previously reported specific tension of human muscle fibers [i.e., 17.1 N/cm2 (43)]. The specific torque is useful to understand the action of each fascicle and muscle portion independent of the size. The medians of moment arm length, force fraction, and specific torque across all fascicles were calculated for each portion.
Based on the 3-D coordinates of fascicles within each portion (anterior-proximal, anterior-distal, and posterior), the AM muscle belly segmentation was divided into three portions (Fig. B in Supplemental Supporting Information S1). Any voxels within AM segmentation that did not contain fascicles (empty voxels) were assigned to the nearest portion, which was determined by calculating the shortest distance between the empty voxel and individual voxels consisting of each portion. If there were more than two portions nearest to the empty voxel, one portion was randomly selected for the assignment. The fat proportion was determined for each portion by averaging the ratio of signal intensity in the fat images to the combined fat and water signal intensities from the Dixon sequence across all voxels of each portion, similar to a previous study (44). The raw muscle volume was calculated for each portion by multiplying voxel size by the number of voxels included in each binary volume and then corrected by the fat proportion in the corresponding portion (fat-corrected muscle volume). The raw and fat-corrected muscle PCSAs were calculated for each portion by dividing the raw muscle volume or fat-corrected muscle volume by the median fascicle length in the corresponding portion, similar to a previous study (45). The potential torque, an estimate of the theoretical maximal torque-generating capacity, was determined for each portion by multiplying the fat-corrected PCSA by the specific torque.
Joint Torque Measurement
The isometric torque during the maximal voluntary contraction (MVC) for the hip extension and adduction was experimentally assessed for the right leg using a dynamometer (Biodex System4, Biodex Medical Systems, Shirley). We did not measure the MVC for the hip internal/external rotation owing to the difficulty in controlling joint configuration during torque measurements.
For the hip extension, the participants sat on the dynamometer seat with the right hip joint at 0° (“neutral” position) or 45° (“flexed” position) flexion, 0° adduction, 0° internal rotation, and the right knee joint at 90° flexion. Their pelvis and thighs were firmly secured by using a nonelastic belt to prevent any rotation of the hip joint during torque measurements. Also, a heavy bed was placed in front of the shank, and an examiner sat on the bed to avoid any extension of the knee by the participants. For the hip adduction, the participants lay on the left side on the dynamometer seat with the hip joint at 0° (“neutral” position) or 45° (“flexed” position) flexion, 0° adduction, 0° internal rotation, and the right knee joint at 90° flexion. Their trunk and pelvis were firmly secured by using nonelastic belts to prevent any rotation of the hip joint or lateral flexion of the trunk. The shank position was fixed by an examiner to prevent any knee extension during torque measurements.
The participants performed several warming-up and familiarization trials of the submaximal isometric hip extension/adduction (three repetitions of 30, 50, and 70% of the subjective maximum). Then, the MVC for the isometric hip extension/adduction for 3 s was performed twice with at least 1 min of rest interval. The participants were instructed to avoid any countermovement at the early phase of torque development. When the maximal joint torque was increased from the first to the second MVCs, further trial(s) were performed until no increase in the maximal joint torque was observed. The joint torque data from the dynamometer was stored in a personal computer via an A/D converter (PowerLab 16SP, ADInstruments, New South Wales, Australia) at a 1,000 Hz sampling rate. For each trial data, the maximal joint torque was determined as the difference between baseline torque and peak torque during the trial. Among the trials, the highest values for maximal joint torque of hip extension/adduction at the neutral and flexed positions were used for further analyses.
Statistics
Data were first assessed for normality using the Shapiro–Wilk test. As some variables (raw muscle volume, fat-corrected muscle volume, fascicle length, moment arm, specific torque, and potential torque) showed a non-normal distribution (P ≤ 0.032), nonparametric tests were used in this study. Differences in fat proportion, raw muscle volume, fat-corrected muscle volume, raw muscle PCSA, fat-corrected muscle PCSA, and fascicle length among the anterior-proximal, anterior-distal, and posterior portions were examined by the Wilcoxon signed rank test with Bonferroni correction. This test was also used to examine differences in moment arm length, force fraction, and specific torque among the hip joint actions (hip extension, adduction, and internal rotation) for each portion and the difference in the whole muscle potential torque between hip extension and adduction. Spearman’s rank correlation coefficients were calculated to investigate relationships between the potential torque of the whole muscle and the corresponding hip adduction/extension torque measured with the dynamometer. The significance level was set at α = 0.05. All statistical analyses were conducted on MATLAB R2023a.
RESULTS
The fat proportion was significantly higher in the posterior (median [interquartile range: IQR]: 9.5% [7.6%–10.0%]) and anterior-proximal (median [IQR]: 10.0% [8%–10.7%]) portions than in the anterior-distal (median [IQR]: 8.7% [7.3%–9.1%]) portions (all P ≤ 0.025; Fig. 3). The raw muscle volume was significantly larger in the posterior portion (median [IQR]: 283.3 [214.7–359.2] cm3) than the anterior-distal portion (median [IQR]: 158.5 [139.2–192.3] cm3), which was in turn significantly larger than the anterior-proximal portion (median [IQR]: 55.3 [49.2–62.0] cm3, all P ≤ 0.003). A similar pattern of variation across portions was observed for the fat-corrected muscle volume (medians [IQR]: 259.6 [197.8–320.0] cm3 for the posterior portion, 144.7 [128.0–175.5] cm3 for the anterior-distal portion, and 50.0 [45.5–56.1] cm3 for the anterior-proximal portion, all P ≤ 0.003) and fascicle length (medians [IQR]: 100 [90–110] mm for the posterior portion, 80 [70–100] mm for the anterior-distal portion, and 60 [50–60] mm for the anterior-proximal portion, all P ≤ 0.005), raw muscle PCSA (medians [IQR]: 25.8 [20.1–35.9] cm2 for the posterior portion, 21.1 (14.2–24.9] cm2 for the anterior-distal portion, and 9.5 [8.4–11.0] cm2 for the anterior-proximal portion, all P ≤ 0.031), and the fat-corrected muscle PCSA (medians [IQR]: 23.6 [18.8–33.1] cm2 for the posterior portion, 19.0 [12.9–23.3] cm2 for the anterior-distal portion, and 8.6 [7.7–9.7] cm2 for the anterior-proximal portion, all P ≤ 0.031).
Figure 3.Fat proportion, raw and fat-corrected muscle volumes, fascicle length, and raw and fat-corrected muscle physiological cross-sectional areas (PCSAs) in the human adductor magnus. Horizontal lines between boxes indicate significant differences (P < 0.05). The symbol “+” indicates an outlier. Ant-dist, anterior-distal portion; Ant-prox: anterior-proximal portion; Post: posterior portion.
The examples of within-muscle distribution of fascicle-specific torque are shown in Fig. 4. The specific torque appeared to vary among individual fascicles and across hip joint actions. The moment arm length was significantly different between the hip joint actions in the posterior (medians [IQR], internal rotation: 69.2 [61.0 to 73.9] mm > extension: 67.2 [59.0 to 72.6] mm > adduction: 20.0 [–12.1 to 30.6] mm, all P ≤ 0.031), anterior-proximal (medians [IQR], adduction: 69.1 [65.9 to 72.8] mm > extension and internal rotation: –60.0 [–62.7 to –59.2] and –34.2 [–35.6 to –30.1] mm, respectively, all P ≤ 0.001), and anterior-distal (medians [IQR], extension: 61.6 [57.7 to 70.5] mm > adduction: 45.9 [37.8 to 63.7] mm > internal rotation: –55.3 [–58.6 to –48.7] mm, all P ≤ 0.001) portions (Fig. 5A). The force fraction was significantly varied across the hip joint actions in the posterior (medians [IQR], extension: 0.98 [0.97–0.98] > adduction: 0.96 [0.96–0.98] > internal rotation: 0.37 [0.35–0.43], all P ≤ 0.010), anterior-proximal (medians [IQR], adduction: 0.97 [0.96–0.98] > internal rotation: 0.94 [0.92–0.95] > extension: 0.43 [0.39–0.45], all P ≤ 0.002), and anterior-distal (medians [IQR], adduction: 0.98 [0.97–0.98] > extension: 0.93 [0.91–0.97] > internal rotation: 0.42 [0.39–0.48], all P ≤ 0.003) portions (Fig. 5B). Also, significant variations in the specific torque were found in the posterior (medians [IQR], extension: 1.12 [0.98 to 1.20] Nm/cm2 > adduction and internal rotation: 0.34 [–0.20 to 0.50] and 0.46 [0.37 to 0.53], respectively, all P < 0.001), anterior-proximal (medians [IQR], adduction: 1.14 [1.08 to 1.20] Nm/cm2 > extension: –0.45 [–0.47 to –0.37] Nm/cm2 > internal rotation: –0.55 [–0.56 to –0.49] Nm/cm2, all P ≤ 0.005), and anterior-distal (medians [IQR], extension: 1.00 [0.92 to 1.11] Nm/cm2 > adduction: 0.77 [0.63 to 1.07] Nm/cm2 > internal rotation: –0.39 [–0.48 to –0.32] Nm/cm2, all P ≤ 0.037) portions (Fig. 5C).
Figure 4.Examples of within-muscle distribution of the specific torque about the hip joint in the human adductor magnus.
Figure 5.Between-plane variation in moment arm length (A), force fraction (B), and specific torque (C). Horizontal lines between boxes indicate significant differences (P < 0.05). The symbol “+” indicates an outlier. Ant-dist, anterior-distal portion; Ant-prox, anterior-proximal portion; Post, posterior portion; rot, rotation.
Figure 6A shows the contribution of each portion to the whole muscle’s potential torque. Briefly, the hip extension potential torque was largely dominated by the posterior portion, followed by the anterior-distal portion. The hip adduction potential torque appeared to be distributed evenly among the portions. Meanwhile, the potential torque for the internal rotation in the posterior portion was offset by those for the external rotation in the other portions. The potential torque of the whole muscle (sum of all portions) was significantly greater for the hip extension than adduction (median [IQR]: 49.8 [25.7 to 65.1] Nm vs. 21.9 [18.7 to 41.2] Nm, P = 0.035; Fig. 6B), whereas the magnitude of the difference varied considerably across individuals (–10.6 to 69.6 Nm; Fig. 6C).
Figure 6.Potential torque for each hip joint action in each portion of the human adductor magnus (A), relationship between hip adduction and extension potential torques of the whole muscle (B), and difference in whole muscle potential torque between hip extension and adduction for each participant (C). Ant-dist, anterior-distal portion; Ant-prox, anterior-proximal portion; F, female; M, male; Post, posterior portion; rot, rotation.
The whole muscle potential torque for the hip extension was significantly correlated with the hip extension exerted torque at the neutral (ρ = 0.614, P = 0.017) and flexed (ρ = 0.736, P = 0.003) positions measured with the dynamometer (Fig. 7). On the contrary, no significant correlations were found between the whole muscle potential torque for the hip adduction and the dynamometer-measured hip adduction torque at neutral (ρ = 0.393, P = 0.148) or flexed (ρ = 0.239, P = 0.389) position.
Figure 7.Relationship between potential torque of the whole adductor magnus at hip neutral position and exerted torque at hip neutral and flexed (45° flexed) positions, which was measured with a dynamometer. The regression line is shown to represent the trend of the significant correlation. ρ and P: Spearman’s rank correlation coefficient and its probability.
DISCUSSION
We revealed a complete picture of the human AM architecture and torque-generating potential in 3-D in living young individuals. The specific torque was greater for the hip extension than adduction in two out of three portions, which covered the majority of the whole muscle. Consequently, the potential torque of the whole muscle was over twofold greater for hip extension than adduction. Furthermore, the architecturally estimated potential torque correlated well with the dynamometer-measured torque for the hip extension but not the adduction. These results support our hypothesis and suggest that AM is designed to primarily act for hip extension rather than adduction, challenging the traditional view of this muscle as a major adductor.
The median fascicle length in the present study was 60 mm for the anterior-proximal portion, 80 mm for the anterior-distal portion, and 100 mm for the posterior portion, whereas that in the previous cadaveric study (11) was 84 mm (“AM1”), 129 mm (mean of “AM2” and “AM3”), and 166 mm (“AM4”). Although the absolute fascicle length differs between the previous and present studies, possibly due to the potential difference in body size, the pattern of variation (posterior > anterior-distal > anterior-proximal) is consistent across the studies. However, the variation in fat-corrected muscle PCSA differs from the previous study [anterior-distal (“AM2” + “AM3”): 11.8 cm2 > posterior (“AM4”): 5.8 cm2 > anterior-proximal (“AM1”): 3.6 cm2, (11)], indicating that the posterior portion in living human AM (46.1% of the total PCSA) is much greater than observed in the cadaveric specimens (26.7% of the total PCSA). This may be due to differences in population [75–91 yr old (11)] and the associated physical activity status between the present and previous studies. Namely, elderly individuals may perform less walking and running, which require large hip extension torque (18), potentially causing selective atrophy of the posterior portion of their AM. Our findings suggest that the posterior portion, together with the anterior-distal portion, accounts for the major force-generating capacity of the whole AM in living young individuals.
The moment arm length in each AM portion considerably varied among hip joint actions, most of which is consistent with the previous study on human cadavers (14). However, compared with the anatomical study (14), both the posterior and anterior-distal portions of AM exhibited a longer moment arm length for hip extension than adduction in this study. Specifically, the anterior-distal (“middle”) portion of AM showed a 37% shorter moment arm length for hip extension than adduction in the anatomical study (14), whereas that in the present study showed a 34% longer moment arm length for hip extension than adduction. The force fraction was weighted toward the hip adduction and internal rotation in the anterior-proximal portion, but the hip extension and adduction in the other portions were almost comparable with each other. Consequently, the specific torque, an estimate of torque-generating capacity for a given cross-sectional area, was greater for the hip extension than adduction in the posterior and anterior-distal portions. This finding, coupled with muscle PCSA distribution, indicates that force generated by the majority of human AM is translated more to the hip extension torque than adduction torque in vivo.
Consequently, the potential torque, a theoretical maximal isometric torque that a muscle can exert, of the whole muscle was greater for the hip extension compared with adduction for most individuals. It has been considered that AM is a prime hip adductor. However, the previous findings come from anatomical observations on a limited number of elderly cadaveric specimens (11, 14). Intramuscular electromyographic data from living human subjects indicated that posterior and anterior portions of AM were more active during maximal isometric torque exertion for hip extension than adduction (16). This was supported by a surface electromyographic finding (17). The observed potential torque, coupled with the previous findings on neuromuscular activation (16, 17), supports that AM primarily acts for hip extension rather than adduction in living humans. This consideration is further supported by the results that positive correlations between the architecturally estimated potential torque and dynamometer-measured maximal torque were found for the hip extension but not for the adduction, which suggests a contribution of AM to hip extension torque generation in human movements. On the contrary, the hip adduction torque provided by AM may not be large enough to explain the interindividual variation of the adduction torque at the hip joint. This may be associated with the previous finding on a lower activity level of AM during hip adduction than extension (16).
It is known that the hip extensors, such as the gluteus maximus, act to stabilize the trunk in the sagittal plane against external disturbance at the foot strike in human running (46). In addition, the angular impulse produced by the hip extensors contributes to most of the forward acceleration of the human body during maximal sprint running (47). Therefore, it is likely that, similar to the gluteus maximus, the AM also plays important roles as a sagittal-plane stabilizer and powerful motor in human locomotion. In fact, human sprint runners are reported to have a large AM as compared with untrained individuals (48). Nevertheless, the potential contribution of AM in human movements has often been overlooked so far. The present study would motivate us to redefine the functional role of human AM and accordingly rename this muscle, for example, “Extensor magnus.”
Interestingly, the difference in AM potential torque between hip extension and adduction appeared to vary greatly across individuals. For example, one showed over 60% greater potential torque for the hip extension compared with adduction, the others had a comparable potential torque between them. These results open the novel possibility that a single muscle mechanical action may not be constant but varies across individuals. The sources of interindividual differences are unclear but may be associated with possible variations in the relative torque-generating capacity of other synergistic muscles. Specifically, individuals with AM showing a smaller potential torque for hip extension may exhibit a greater potential torque of other synergists such as the gluteus maximus to meet the demands of hip extension torque exertion in motor tasks. Also, it is widely known that human males and females exhibit different morphological characteristics of the pelvis (49) and femur (50), which are attachments of AM. However, we did not find any clear sex-related differences in potential torque balance (extension vs. adduction; Fig. 6C), which are worth examining in future studies with a larger sample size.
The architectural parameters, including moment arm, force fraction, and muscle PCSA, are usually incorporated into musculoskeletal simulation models to study the mechanisms underlying human and animal movements (51, 52). However, the generic model in such a simulation normally stems from anatomical studies on human cadavers. Charles et al. (53) reported that human lower limb models derived from muscle architectural parameters in each individual improved the accuracy of the simulation. The present findings on large individual variations of architecturally estimated muscle action also highlight the importance of subject-specific modeling in musculoskeletal simulation.
The major limitation of this study is that we measured AM architecture and potential action at the hip neutral position, although they can vary depending on the joint configuration. In this regard, we simulated how the moment arm length, force fraction, and specific torque change across a range of hip joint angles by applying the rotation around the hip extension-flexion axis to the representative line of action vector defined by the mean position coordinates of fascicle origins and midpoints in each portion (Fig. A in Supplemental Supporting Information S2). The results showed that specific torques in the posterior and anterior-distal portions were almost always greater for hip extension than adduction across the range of hip flexion. Notably, those specific torques in the anterior-proximal portion became comparable at around 30° hip flexion. These results support that AM is designed to primarily act for hip extension rather than adduction. However, we cannot rule out the effect of fascicle curvature along the length and its change during movements on the mechanical action estimates. The joint torques were measured in the knee-flexed position, whereas AM potential torques were estimated from MR images acquired in the knee-extended position. Because AM is a mono-articular muscle spanning the hip joint only, its architecture is not largely affected by knee joint angle. However, the knee joint angle may affect the relative contribution of AM among synergistic muscles to hip joint torque, as some synergists, such as the hamstrings, span across the knee joint and may change their torque-generating potential. Therefore, further comprehensive examinations matching joint configuration between measured and estimated torques are needed to understand how AM contributes to hip joint performance. Also, we measured hip joint torque during voluntary contractions and did not measure physiologically maximal joint torque for each individual. Furthermore, although we divided AM into three portions, the anterior-distal portion can show regional architectural variation along the femur because of its broad insertion. However, even when we divided the anterior-distal portion into two subportions, we found similar results that the whole AM showed a much greater potential torque for hip extension than adduction (46.0 Nm vs. 23.1 Nm, P = 0.048). This suggests that the pattern of compartmentalization within the AM does not have a significant influence on the present findings. Also, we conducted a sensitivity analysis to test how tracking parameters (maximal turning angle, step size, and the number of tracked fibers) setting affects our findings. The results showed that changes in the maximal turning angle and step size resulted in changes in potential torque estimates (Supplemental Supporting Information S3). However, our main result that the hip extension potential torque of AM was greater than the adduction torque was not changed across the different parameter settings. Meanwhile, we recruited a small number of healthy young adults in this study, and the generalizability of our findings across populations with different backgrounds (e.g., sex and age) remains unclear. Further studies examining individual variation and plasticity of muscle 3-D architecture and potential action in relation to mechanical loading/unloading and pathology would extend our understanding of muscle functional roles and hint at training/rehabilitation strategies to effectively improve human motor performance in the future.
Conclusions
We revealed a comprehensive picture of the human AM architecture and torque-generating potential in 3-D in living young individuals. The specific torque was greater for hip extension than adduction in the posterior and anterior-distal portions covering the majority of the whole muscle volume and PCSA. Consequently, the potential torque of the whole muscle was much greater for hip extension than adduction. Furthermore, the architecturally estimated potential torque correlated well with the dynamometer-measured exerted torque for hip extension but not for adduction. These results strongly suggest that the human AM is designed to primarily act for hip extension rather than adduction in living young individuals, challenging the traditional view of this giant muscle as a major adductor.
DATA AVAILABILITY
Data will be made available upon reasonable request.
SUPPLEMENTAL MATERIAL
Supplemental Supporting Information S1–S3: https://doi.org/10.6084/m9.figshare.28429493.v1.
GRANTS
This study was supported by JSPS KAKENHI, Grant Number JP23K19943.
DISCLOSURES
No conflicts of interest, financial or otherwise, are declared by the authors.
AUTHOR CONTRIBUTIONS
K.T. and T.W. conceived and designed research; K.T. and H.T. performed experiments; K.T. analyzed data; K.T., H.T., R.K., and T.W. interpreted results of experiments; K.T. prepared figures; K.T. drafted manuscript; H.T., R.K., and T.W. edited and revised manuscript; K.T., H.T., R.K., and T.W. approved final version of manuscript.
ACKNOWLEDGMENTS
We acknowledge our laboratory members for helpful comments on this study.
REFERENCES
- 1. . Dimensions and moment arms of the hind- and forelimb muscles of common chimpanzees (Pan troglodytes). Am J Phys Anthropol 110: 179–199, 1999. doi:10.1002/(SICI)1096-8644(199910)110:2<179::AID-AJPA5>3.0.CO;2-Z.
Crossref | PubMed | Web of Science | Google Scholar - 2. . Endurance running and the evolution of Homo. Nature 432: 345–352, 2004. doi:10.1038/nature03052.
Crossref | PubMed | Web of Science | Google Scholar - 3. . Morphological analysis of the hindlimb in apes and humans. I. Muscle architecture. J Anat 208: 709–724, 2006. doi:10.1111/j.1469-7580.2006.00563.x.
Crossref | PubMed | Web of Science | Google Scholar - 4. . Body composition in Pan paniscus compared with Homo sapiens has implications for changes during human evolution. Proc Natl Acad Sci USA 112: 7466–7471, 2015. doi:10.1073/pnas.1505071112.
Crossref | PubMed | Web of Science | Google Scholar - 5. . Are current measurements of lower extremity muscle architecture accurate? Clin Orthop Relat Res 467: 1074–1082, 2009. doi:10.1007/s11999-008-0594-8.
Crossref | PubMed | Web of Science | Google Scholar - 6. . Relationships of 35 lower limb muscles to height and body mass quantified using MRI. J Biomech 47: 631–638, 2014. doi:10.1016/j.jbiomech.2013.12.002.
Crossref | PubMed | Web of Science | Google Scholar - 7. . Groin injuries in sports medicine. Sports Health 2: 231–236, 2010. doi:10.1177/1941738110366820.
Crossref | PubMed | Google Scholar - 8. . Human balance and posture control during standing and walking. Gait Posture 3: 193–214, 1995. doi:10.1016/0966-6362(96)82849-9.
Crossref | Google Scholar - 9. . Explaining the hip adduction moment variability during gait: implications for hip abductor strengthening. Clin Biomech (Bristol) 24: 267–273, 2009. doi:10.1016/j.clinbiomech.2008.12.006.
Crossref | PubMed | Web of Science | Google Scholar - 10. . Effect of running speed on lower limb joint kinetics. Med Sci Sports Exerc 43: 1260–1271, 2011. doi:10.1249/MSS.0b013e3182084929.
Crossref | PubMed | Web of Science | Google Scholar - 11. . Why adductor magnus muscle is large: the function based on muscle morphology in cadavers. Scand J Med Sci Sports 24: 197–203, 2014. doi:10.1111/j.1600-0838.2012.01466.x.
Crossref | PubMed | Web of Science | Google Scholar - 12. . The adductor magnus “mini-hamstring”: MRI appearance and potential pitfalls. Skeletal Radiol 45: 213–219, 2016. doi:10.1007/s00256-015-2291-5.
Crossref | PubMed | Web of Science | Google Scholar - 13. . Anatomy of the adductor magnus origin: implications for proximal hamstring injuries. Orthop J Sports Med 4:
2325967115625055 , 2016. doi:10.1177/2325967115625055.
Crossref | PubMed | Web of Science | Google Scholar - 14. . Actions of hip muscles. Phys Ther 66: 351–361, 1986. doi:10.1093/ptj/66.3.351.
Crossref | PubMed | Web of Science | Google Scholar - 15. . Effect of aging on human muscle architecture. J Appl Physiol (1985) 95: 2229–2234, 2003. doi:10.1152/japplphysiol.00433.2003.
Link | Web of Science | Google Scholar - 16. . Adductor magnus: an EMG investigation into proximal and distal portions and direction-specific action. Clin Anat 31: 535–543, 2018. doi:10.1002/ca.23068.
Crossref | PubMed | Web of Science | Google Scholar - 17. . Effect of hip angle on neuromuscular activation of the adductor longus and adductor magnus muscles during isometric hip flexion and extension. Eur J Appl Physiol 119: 1611–1617, 2019. doi:10.1007/s00421-019-04150-5.
Crossref | PubMed | Web of Science | Google Scholar - 18. . Modulation of work and power by the human lower-limb joints with increasing steady-state locomotion speed. J Exp Biol 218: 2472–2481, 2015. doi:10.1242/jeb.119156.
Crossref | PubMed | Web of Science | Google Scholar - 19. . Adductor magnus: extending the knowledge–a short review of structure and function. Int J Osteopath Med 49:
100671 , 2023. doi:10.1016/j.ijosm.2023.100671.
Crossref | Web of Science | Google Scholar - 20. . Physiological cross-sectional area of human leg muscles based on magnetic resonance imaging. J Orthop Res 10: 928–934, 1992. doi:10.1002/jor.1100100623.
Crossref | PubMed | Web of Science | Google Scholar - 21. . Muscle-fiber pennation angles are greater in hypertrophied than in normal muscles. J Appl Physiol (1985) 74: 2740–2744, 1993. doi:10.1152/jappl.1993.74.6.2740.
Link | Web of Science | Google Scholar - 22. . In vivo human gastrocnemius architecture with changing joint angle at rest and during graded isometric contraction. J Physiol 496: 287–297, 1996. doi:10.1113/jphysiol.1996.sp021685.
Crossref | PubMed | Web of Science | Google Scholar - 23. . Changes in Achilles tendon moment arm from rest to maximum isometric plantarflexion: in vivo observations in man. J Physiol 510: 977–985, 1998. doi:10.1111/j.1469-7793.1998.977bj.x.
Crossref | PubMed | Web of Science | Google Scholar - 24. . Architectural and functional features of human triceps surae muscles during contraction. J Appl Physiol (1985) 85: 398–404, 1998. doi:10.1152/jappl.1998.85.2.398.
Link | Web of Science | Google Scholar - 25. . Intra- and intermuscular variation in human quadriceps femoris architecture assessed in vivo. J Anat 209: 289–310, 2006. doi:10.1111/j.1469-7580.2006.00619.x.
Crossref | PubMed | Web of Science | Google Scholar - 26. . Three-dimensional architecture of human medial gastrocnemius fascicles in vivo: regional variation and its dependence on muscle size. J Anat 241: 1324–1335, 2022. doi:10.1111/joa.13750.
Crossref | PubMed | Web of Science | Google Scholar - 27. . Coronal as well as sagittal fascicle dynamics can bring about a gearing effect in muscle elongation by passive lengthening. Med Sci Sports Exerc 55: 2035–2044, 2023. doi:10.1249/MSS.0000000000003229.
Crossref | PubMed | Web of Science | Google Scholar - 28. . Changes in lower limb muscle cross-sectional area and tissue fluid volume after transition from standing to supine. Acta Physiol Scand 148: 379–385, 1993. doi:10.1111/j.1748-1716.1993.tb09573.x.
Crossref | PubMed | Web of Science | Google Scholar - 29. . History-dependence of muscle slack length in humans: effects of contraction intensity, stretch amplitude, and time. J Appl Physiol (1985) 129: 957–966, 2020. doi:10.1152/japplphysiol.00106.2020.
Link | Web of Science | Google Scholar - 30. . End-divergent architecture diversifies within-muscle mechanical action in human gluteus maximus in vivo. J Biomech 179:
112488 , 2025. doi:10.1016/j.jbiomech.2024.112488.
Crossref | PubMed | Web of Science | Google Scholar - 31. . 3D slicer as an image computing platform for the quantitative imaging network. Magn Reson Imaging 30: 1323–1341, 2012. doi:10.1016/j.mri.2012.05.001.
Crossref | PubMed | Web of Science | Google Scholar - 32. . Deterministic diffusion fiber tracking improved by quantitative anisotropy. PLoS One 8:
e80713 , 2013. doi:10.1371/journal.pone.0080713.
Crossref | PubMed | Web of Science | Google Scholar - 33. . Diffusion weighted image denoising using overcomplete local PCA. PLoS One 8:
e73021 , 2013. doi:10.1371/journal.pone.0073021.
Crossref | PubMed | Web of Science | Google Scholar - 34. . Reliability and robustness of muscle architecture measurements obtained using diffusion tensor imaging with anatomically constrained tractography. J Biomech 86: 71–78, 2019. doi:10.1016/j.jbiomech.2019.01.043.
Crossref | PubMed | Web of Science | Google Scholar - 35. . Three-dimensional architecture of the human subscapularis muscle in vivo. J Biomech 161:
111854 , 2023. doi:10.1016/j.jbiomech.2023.111854.
Crossref | PubMed | Web of Science | Google Scholar - 36. . Validation of diffusion tensor MRI-based muscle fiber tracking. Magn Reson Med 48: 97–104, 2002. doi:10.1002/mrm.10198.
Crossref | PubMed | Web of Science | Google Scholar - 37. . Determination of three-dimensional muscle architectures: validation of the DTI-based fiber tractography method by manual digitization. J Anat 223: 61–68, 2013. doi:10.1111/joa.12062.
Crossref | PubMed | Web of Science | Google Scholar - 38. . Morphologic and morphometric characteristics of the adductor minimus in Korean: its topographic relationship with respect to neighbouring anatomical structure and clinical significance. Folia Morphol (Warsz) 80: 963–971, 2021. doi:10.5603/FM.a2020.0111.
Crossref | PubMed | Web of Science | Google Scholar - 39. . Prediction of the hip joint centre in adults, children, and patients with cerebral palsy based on magnetic resonance imaging. J Biomech 40: 595–602, 2007. doi:10.1016/j.jbiomech.2006.02.003.
Crossref | PubMed | Web of Science | Google Scholar - 40. . The lower limbs of sprinters have larger relative mass but not larger normalized moment of inertia than controls. Med Sci Sports Exerc 55: 590–600, 2023. doi:10.1249/MSS.0000000000003064.
Crossref | PubMed | Web of Science | Google Scholar - 41. ;
Standardization and Terminology Committee of the International Society of Biomechanics. ISB Recommendation on Definitions of joint coordinate system of various joints for the reporting of human joint motion–part I: ankle, hip, and spine. J Biomech 35: 543–548, 2002. doi:10.1016/s0021-9290(01)00222-6.
Crossref | PubMed | Web of Science | Google Scholar - 42. . In vivo determination of the Achilles tendon moment arm in three-dimensions. J Biomech 45: 409–413, 2012. doi:10.1016/j.jbiomech.2011.10.018.
Crossref | PubMed | Web of Science | Google Scholar - 43. . Direct intraoperative measurement of isometric contractile properties in living human muscle. J Physiol 601: 1817–1830, 2023. doi:10.1113/JP284092.
Crossref | PubMed | Web of Science | Google Scholar - 44. . Quantity versus quality: age-related differences in muscle volume, intramuscular fat, and mechanical properties in the triceps surae. Exp Gerontol 156:
111594 , 2021. doi:10.1016/j.exger.2021.111594.
Crossref | PubMed | Web of Science | Google Scholar - 45. . Regional muscle features and their association with knee extensors force production at a single joint angle. Eur J Appl Physiol 123: 2239–2248, 2023. doi:10.1007/s00421-023-05237-w.
Crossref | PubMed | Web of Science | Google Scholar - 46. . The human gluteus maximus and its role in running. J Exp Biol 209: 2143–2155, 2006. doi:10.1242/jeb.02255.
Crossref | PubMed | Web of Science | Google Scholar - 47. . Lower-limb joint mechanics during maximum acceleration sprinting. J Exp Biol 222:
jeb209460 , 2019. doi:10.1242/jeb.209460.
Crossref | PubMed | Web of Science | Google Scholar - 48. . Gluteus and posterior thigh muscle sizes in sprinters: their distributions along muscle length. Eur J Sport Sci 22: 799–807, 2022. doi:10.1080/17461391.2021.1907450.
Crossref | PubMed | Web of Science | Google Scholar - 49. . The human pelvis: variation in structure and function during gait. Anat Rec (Hoboken) 300: 633–642, 2017. doi:10.1002/ar.23552.
Crossref | PubMed | Web of Science | Google Scholar - 50. . Gender differences in 3D morphology and bony impingement of human hips. J Orthop Res 29: 333–339, 2011. doi:10.1002/jor.21265.
Crossref | PubMed | Web of Science | Google Scholar - 51. . OpenSim: open-source software to create and analyze dynamic simulations of movement. IEEE Trans Biomed Eng 54: 1940–1950, 2007. doi:10.1109/TBME.2007.901024.
Crossref | PubMed | Web of Science | Google Scholar - 52. . OpenSim: simulating musculoskeletal dynamics and neuromuscular control to study human and animal movement. PLoS Comput Biol 14:
e1006223 , 2018. doi:10.1371/journal.pcbi.1006223.
Crossref | PubMed | Web of Science | Google Scholar - 53. . Subject-specific muscle properties from diffusion tensor imaging significantly improve the accuracy of musculoskeletal models. J Anat 237: 941–959, 2020. doi:10.1111/joa.13261.
Crossref | PubMed | Web of Science | Google Scholar
