Association Between Lumbopelvic Stability During a Single-Legged Step Down and Elbow-Varus Torque During Baseball Pitching

Key Points

• Higher elbow torque was separately associated with increased transverse-plane pelvis and trunk motion during a single-legged step-down (SLSD) task.

• The SLSD is a simple and repeatable test to estimate pitchers’ risk of increased elbow-varus torque during pitching.

• Clinicians can assess trunk and pelvis kinematics during the SLSD to evaluate throwing-elbow injury risk and performance.

Ulnar collateral ligament (UCL) sprains of the elbow are common in baseball pitchers.¹ The UCL is a passive restraint stabilizing the medial elbow, resisting approximately 50% of elbow-valgus torque during pitching.^2,3 High (internal) elbow-varus torque during the pitch has been implicated in UCL injury.⁴ One factor associated with increased elbow-varus torque during pitching is high ball velocity, but this relationship is inconsistent across pitchers.^5–7 Some pitchers can throw faster without a meaningful increase in elbow-varus torque, whereas others have a strong torque-velocity (T-V) relationship. Differences in physical capacities such as joint stability and control can mediate the T-V relationship, contributing to the increased variance observed among pitchers.⁸

The lumbopelvic junction is the intersection through which energy is transferred from the lower extremities to the trunk and then to the upper extremity during a baseball pitch.^9–11 A stable base at the lumbopelvic junction is important for both trunk and upper extremity motion during the pitch.¹² Increased stability and control at the lumbopelvic junction are thought to facilitate sequenced rotation of superior segments through improved segment-rotation timing and control of trunk rotation.^12,13 Lumbopelvic stability has been associated with the magnitude of elbow-varus torque and pitching performance.^8,14,15

Deficits in lumbopelvic stability in baseball pitchers have been detected during dynamic single-legged tasks, such as the single-legged step-down (SLSD), single-legged squat, or single-legged balance task. Specifically, increased trunk lateral tilt and increased pelvic tilt in the sagittal plane can predict increases in elbow-varus torque, elbow injury rates, and associated time loss.^8,13,16 Observations during such tasks can predict kinematics during the pitch. For instance, increased trunk lean observed during the single-legged squat is associated with the same deviation during the baseball pitch.^17,18 However, the SLSD is a relatively simple task that, on its own, may not be challenging enough for trained, athletic participants to show differences in lumbopelvic stability. To address this concern, we aimed to advance previous research by using a Stroop condition during the SLSD to increase the ecological validity of the lumbopelvic stability assessment. A Stroop condition uses an auditory and cognitive element to mimic stimuli that demand external focus during the pitch. This condition may define more nuanced differences in lumbopelvic stability. It remains unclear whether lumbopelvic stability assessed during an SLSD predicts elbow-varus torque during pitching.

The primary purpose of our study was to determine the predictive ability of lumbopelvic stability assessed during the SLSD task on elbow-varus torque while controlling for ball velocity. We hypothesized that increased triplanar motion during an SLSD task would predict higher elbow-varus torque and lower ball velocity. To accomplish this, we first identified distinct clusters of pitchers characterized by elbow torque and ball velocity. Second, we determined how triplanar lumbopelvic kinematic variables during the SLSD (non-Stroop and Stroop) formed constructs of lumbopelvic stability. Third, we determined the relationship between lumbopelvic stability constructs and the pitcher clusters (subgroups). Characterizing how lumbopelvic stability relates to elbow torque can inform efforts aimed at reducing elbow injury during pitching.

METHODS

Procedures

Participants completed 2 assessments within 48 hours of each other: an SLSD to assess lumbopelvic stability and a pitching bout to assess elbow-varus torque and ball velocity. For the pitching assessment, pitchers first completed their typical warm-up, which consisted of mobility, banded arm-care exercises, and long toss. Pitchers then threw 10 fastballs, 5 at 75% effort and 5 at 100% effort, from a regulation-size and -distance mound to a catcher. Elbow-varus torque was collected using a single inertial measurement unit secured to the ulna 2 finger widths distal to the medial humeral epicondyle with tape and self-adhering wrap (model PULSE; Driveline Baseball). Ball velocity was measured using a Stalker Sport 2 radar gun (Stalker Radar) positioned behind home plate. Test-retest reliability for elbow-varus torque was excellent (intraclass correlation coefficient [3,2] = 0.94), and error was low (standard error of the measurement = 3.4 N·m).

Pitchers performed an SLSD from a 20-cm box for lumbopelvic stability assessment. Six wireless inertial measurement units (MTw Awinda; Xsens) were secured to the participant using double-sided tape and hook-and-loop wraps on the following landmarks: bilaterally on the lateral border of the shank 2 finger widths below the tibial plateau, bilaterally on the lateral border of the thigh 4 finger widths above the lateral femoral epicondyle, on the sacrum between the posterior and superior iliac spines, and over the sternum on top of the xiphoid process (Figure 1). We ensured proper magnetometer function to enable sensor fusion correction. Sensors were calibrated with participants in a neutral posture and facing magnetic north, and the collection area was positioned 15 to 20 feet (4.6–6.1 m) from any other electronic devices to limit magnetic destabilization. The standardized neutral calibration position required participants to stand in line with an arrow while trunk position was standardized using the following command: “Slouch your upper body, now stand up tall like you are in the military, now relax by 10% and hold that position.” Standardized instructions for the SLSD and practice time were given to participants. For both normal and Stroop conditions, participants performed 1 set of 10 repetitions on each lower extremity. A repetition began with participants standing on the 20-cm box with their hands crossed over their shoulders and the noninvolved lower extremity hanging extended in front, off the box. Next, they stepped down until the opposite heel touched the floor, then returned to the original position (Figure 2). Additional instructions were provided before completion of the Stroop condition. An audio recording of a random combination of the words high and low was played, and the 2 words were played randomly at either a high pitch or low pitch. During the SLSD, participants had to orally indicate the pitch of the spoken word, regardless of the stated word. Participants practiced the SLSD with the auditory Stroop before data collection. To limit accelerometer and gyroscopic drift, we limited the duration of each data collection to approximately 5 minutes per participant. Test-retest reliability of lumbopelvic kinematics conducted during the SLSD indicated excellent reliability (intraclass correlation coefficient [3,2] = 0.83–0.99), with low error (standard error of the measurement = 0.25°–1.19°).

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

RESULTS

A total of 44 male pitchers (age = 19.6 ± 1.3 years, height = 190 ± 10 cm, mass = 90.1 ± 6.3 kg) participated. Pitching torque or velocity variables were missing for 6 participants, leaving 38 participants for data analysis. We performed logistic regression on combined pitching and SLSD data. Mean torque was 55.6 ± 11.5 N·m, and mean ball velocity was 133.3 km/hr (82.9 ± 42.8 mph).

Cluster Analysis

A 2-cluster solution (silhouette criterion of 0.50) separated pitchers into 2 subgroups: cluster 1 (n = 17) and cluster 2 (n = 21) (Figures 3 and 4). Members of cluster 1 (low torque–high velocity) had a lower torque (48.3 ± 11.1 N·m vs 61.6 ± 8.3 N·m; P < .001) and higher ball velocity [136 + 3.2 km/hr (84.8 ± 2.0 mph vs 130.8 + 3.8 km/hr (81.3 ± 2.4 mph); P < .001] compared with members of cluster 2 (high torque–low velocity).

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

Principal Components Analysis

We entered a total of 12 variables into each PCA: sagittal-plane lead-leg non-Stroop and Stroop, sagittal-plane trail-leg non-Stroop and Stroop, frontal-plane lead-leg non-Stroop and Stroop, frontal-plane trail-leg non-Stroop and Stroop, transverse-plane lead-leg non-Stroop and Stroop, and transverse-plane trail-leg non-Stroop and Stroop (Table 1). For both trunk and pelvis SLSD variables, sample adequacy using the Kaiser-Meyer-Olkin test (0.65 and 0.68, respectively) and suitability for data reduction using the Bartlett test of sphericity (all P < .001) were confirmed. For the trunk, only 4 components were extracted, for a total explained 85.4% variance (Table 2): (1) PC 1: sagittal plane (eigenvalue = 4.0, 33.7% variance, 4 variables), (2) PC 2: transverse plane (eigenvalue = 3.5, 29.3% variance, 4 variables), (3) PC 3: frontal-plane trail leg (eigenvalue = 1.5, 12.1% variance, 2 variables), and (4) PC 4: frontal-plane lead leg (eigenvalue = 1.2, 10.2% variance, 2 variables). Factor scores consisted of items loading at ≥0.820 on a component (Table 3), with all 12 trunk SLSD variables included in the final PCA solution. For the pelvis SLSD variables, only 4 components were extracted and, in total, explained 90.0% of the variance (Table 2): (1) PC 1: sagittal plane (eigenvalue = 5.2, 43.6% variance, 4 variables), (2) PC 2: transverse plane (eigenvalue = 2.1, 17.5% variance, 4 variables), (3) PC 3: frontal-plane trail leg (eigenvalue = 1.9, 16.2% variance, 2 variables), and (4) PC 4: frontal-plane lead leg (eigenvalue = 1.5, 12.7% variance, 2 variables). Factor scores consisted of items loading at ≥0.867 on a component (Table 4). We included all 12 pelvis SLSD variables (Table 1) in the final PCA solution (Table 2). In summary, based on the PCA findings, we developed a classification system for the SLSD based on the plane of motion and limb type (relevant for the frontal plane only). This system revealed 4 dominant patterns of variability in pitchers’ performance on the SLSD: transverse plane, sagittal plane, frontal-plane trail leg, and frontal-plane lead leg.

Table 1. Descriptive Data for Single-Legged Step Down (Mean ± SD)^a

View Fullscreen

Table 2. Summary of Variables Captured by Principal Components Analysis for the Trunk and Pelvis

View Fullscreen

Table 3. Trunk Principal Components (PC) Analysis Loading Matrix for Single-Legged Step-Down Kinematics

View Fullscreen

Table 4. Pelvis Principal Components (PC) Analysis Loading Matrix for Single-Legged Step-Down Kinematics

View Fullscreen

Logistic Regression

Logistic models predicted the high torque–low velocity cluster using the SLSD kinematic variables’ PCs (Tables 5 and 6). For the trunk variables, only PC 2: transverse plane was different (odds ratio [OR] = 2.9; 95% CI = 1.1, 7.7; P = .03), with higher PC 2: transverse plane values for the trunk predicting higher odds of belonging to the high torque–low velocity pitcher cluster subgroup. Higher PC 2: transverse plane values for the trunk corresponded to higher transverse-plane motion for all underlying metrics (Figure 5). Therefore, higher transverse-plane motion for the lead and trail legs (Stroop and non-Stroop conditions) for the trunk predicted higher odds of belonging to the high torque–low velocity pitcher cluster subgroup. The overall fit (R²) was 26% (model 1 in Table 5). After the stepwise procedure, we entered weight, which increased the model fit to 42%. With the inclusion of weight, PC 2: transverse plane remained different (OR = 2.9; 95% CI = 1.1, 8.0; P = .04) (model 2 in Table 5).

Table 5. Trunk Principal Components (PC) Order by Odds Ratios for Multivariate Logistic Regression to Predict High Torque–Low Velocity Cluster Membership

View Fullscreen

Table 6. Pelvis Principal Components (PC) Order by Odds Ratios for Multivariate Logistic Regression to Predict High Torque–Low Velocity Cluster Membership

View Fullscreen

View Figure

• Download as Powerpoint
• Download Figure

For the pelvis variables, only PC 2: transverse plane was different (OR = 2.2; 95% CI = 1.0, 4.8; P = .049), with higher PC 2: transverse plane pelvis values predicting higher odds of belonging to the high torque–low velocity pitcher cluster. Higher PC 2: transverse plane values for the pelvis corresponded to higher transverse-plane motion for all underlying metrics (Figure 6). Therefore, higher transverse-plane motion for the lead and trail legs (Stroop and non-Stroop conditions) for the pelvis predicted higher odds of belonging to the high torque–low velocity pitcher cluster subgroup. The overall model fit (R²) was 25% (model 1 in Table 6). After the stepwise procedure, we also entered weight, increasing the model fit to 39%. With the inclusion of weight, PC 2: transverse plane remained different (OR = 2.5; 95% CI = 1.1, 6.0; P = .03) (model 2 in Table 6).

View Figure

• Download as Powerpoint
• Download Figure

DISCUSSION

Proximal stability is generally considered to improve distal mobility.¹² Within the framework of pitching, the ability to minimize unwanted lumbopelvic movement is thought to facilitate sequenced rotation of superior segments through improved segment rotation timing.^12,13 Increased lumbopelvic stability in pitchers has been associated with decreased upper extremity demand, lower UCL injury risk, and improved pitching performance.^8,13,14,16 In this study, we aimed to determine if lumbopelvic stability during the SLSD could predict elbow-varus torque while controlling for ball velocity. Our results partially confirmed our hypothesis that triplanar kinematics would predict the T-V relationship. Specifically, increased transverse-plane trunk and pelvis motion during the SLSD separately predicted membership in the high torque–low velocity cluster subgroup. To the best of our knowledge, we are the first to use unsupervised methods to identify clusters within the T-V relationship, as well as the first to quantify dynamic lumbopelvic-stability–associated T-V clusters in baseball pitchers. By clarifying the elements of lumbopelvic stability in an SLSD task that best differentiate the cluster subgroups of low torque–high velocity versus high torque–low velocity, this research provides knowledge that can inform efforts to reduce UCL injury.

We are the first to successfully use unsupervised methods to identify subgroups in the T-V relationship. Due to limited subgroup sizes, we were limited to testing 2- and 3-subgroup solutions. However, the 2-cluster solution achieved a silhouette criterion of 0.50 (good), indicating well-matched and tightly fit groups. Members of the high torque–low velocity group had higher elbow-varus torque and lower ball velocity compared with members in the low-torque–high-velocity group (Figure 4). Given a larger sample size, different subgroups may emerge. Future work should be done to independently validate the cluster groups.

Clinical paradigms are often used at the expense of ecological validity to measure physical capacities such as lumbopelvic stability. The relative simplicity of these tasks facilitates clearer, more specific findings about the physical capacity in question. Despite their decreased ecological validity, findings from clinical paradigms like the SLSD are often reflected in the mechanics of athletic movement outside of the task itself, and pitching is no exception.^17,18,21 These simpler movements can be paired with dual-tasking paradigms to improve the ecological validity of a task by mimicking the in-game cognitive demands of the sport, increasing the difficulty and consequently movement variability.²² Separate PCAs for the pelvis and trunk identified 4 domains within the kinematic data across the non-Stroop and Stroop conditions: (1) sagittal-plane kinematics, (2) transverse-plane kinematics, (3) frontal-plane kinematics for the trail leg, (4) and frontal-plane kinematics for the lead leg. Interestingly, only increased transverse-plane trunk and pelvis rotation during the SLSD were separate predictors of high-torque–low-velocity cluster membership. Specifically, pitchers with increased transverse-plane trunk movement were 2.9 times more likely to be a member of the high-torque–low-velocity group after accounting for body weight. Increased transverse-plane pelvis movement similarly increased high-torque–low-velocity membership likelihood by 2.5 times after accounting for body weight. Although transverse-plane trunk and pelvis motion during the SLSD are generally considered to be compensatory mechanics, they can also be a byproduct of SLSD task performance.²³ Participants were not told how to complete the test; rather, they were instructed to lower themselves to the ground until the opposite heel touched the floor. Increased transverse-plane trunk and pelvis rotation, as well as increased frontal-plane leg movement, may reflect a strategy to achieve heel contact with the floor rather than true compensatory movement.

Gluteal muscle performance controls multiplanar pelvis stability and thus may underlie the increased transverse-plane motion seen in the high-torque–low-velocity group. During the baseball pitch, trail-leg gluteal muscles influence transverse-plane pelvis rotation and trunk-pelvis dissociation, in addition to stabilizing extraplanar movement.^9,11,24 The lead-leg gluteal muscles are tasked with stabilizing the pelvis during the late arm-cocking and arm-acceleration phases. Immediately after lead-foot contact, the lead-leg gluteus maximus and medius experience high levels of activity as they absorb energy to keep the pelvis level and decelerate the anterior and inferior translation of the center of mass in unipedal stance.²⁵ Both the gluteus maximus and medius work to provide lumbopelvic stability through resistance to hip flexion, internal rotation, and pelvis lateral tilt during arm acceleration.²⁶ The SLSD offers similar lumbopelvic stabilization demands in that proper execution of the task requires minimization of unwanted, extraplanar movement while descending in single-legged stance. Researchers have demonstrated that individuals with sufficient gluteal muscle strength and coordination have decreased pelvis movement, variability, and compensatory kinematics in the trunk and lower extremities during pitching.^27,28

The core muscles function at the lumbopelvic junction, the anatomic link between the pelvis and trunk. During the pitch, the core and gluteal muscles stabilize the pelvis and control trunk axial rotation, affecting ball velocity and upper extremity moments that include elbow-varus torque.¹¹ Although we analyzed the trunk and pelvis separately, Lewis et al, analyzing SLSD kinematics, showed that transverse-plane pelvis and trunk rotation tend to occur in the same direction.²³ These aligned and concurrent rotations suggest trunk kinematics are related to proximal stability at the pelvis. During the baseball pitch, sequential transverse-plane rotations of the pelvis and trunk form the primary energy delivery mechanism between the lower and upper extremities.²⁹ A pelvis that is not adequately stabilized or rotated may compromise energy transfer to the trunk. Lumbopelvic stability may be improved with increased gluteal and core muscle performance and thus convert a high-torque–low-velocity pitcher who displays poor transverse-plane trunk and pelvis control during an SLSD into a low-torque–high-velocity pitcher.

Limitations of our results include the use of a single inertial sensor to measure elbow-varus torque and the need for validation of our clustering approach. Although the Motus pulse IMU (inertial measurement unit) has demonstrated good to excellent reliability,^30-33 and research supports the use of the Motus pulse IMU for measuring elbow kinetics, with authors reporting a moderate to strong relationship between IMU-derived elbow torque and motion-capture–derived elbow torque,^30,31 validation studies indicate that the IMU can underestimate the magnitude of elbow-varus torque compared to motion-capture methods.^30,34 Therefore, comparison of values derived from these 2 methods is not recommended. Hence, we have not compared elbow-varus torque values from our study with others derived from motion capture. Additionally, while we defined a 2-cluster solution to group the T-V relationship as high or low torque, these clusters and definitions need to be validated in other samples of pitchers to ensure their robustness. Despite these limitations, our results provide coaches and clinicians with a clinical paradigm of the SLSD to assess lumbopelvic stability and gauge a pitcher’s risk of increased elbow-varus torque for a given ball velocity. Pitching biomechanics have been related to SLSD mechanics,^17,35 but the underlying mechanisms of this relationship have not been identified. We identified that increased transverse-plane trunk and pelvis kinematics during an SLSD were separately related to pitchers with high elbow torque during a baseball pitch.

CONCLUSIONS

Lumbopelvic stability during the SLSD can predict elbow-varus torque while controlling for ball velocity during a baseball pitch. Specifically, increased transverse-plane trunk and pelvis motion, while controlling for body weight, separately predicted membership in the high torque–low velocity group. The T-V clusters of high torque–low velocity and low torque–high velocity were identified using unsupervised learning. Plane of motion, rather than SLSD condition (non-Stroop versus Stroop), described variance in SLSD kinematics. Pitchers with increased transverse-plane pelvis or increased trunk motion during an SLSD are at higher risk for higher elbow-varus torque and concurrent lower ball velocity during the baseball pitch. These results afford clinicians with simple criteria for assessing trunk and pelvis kinematics during the SLSD, with implications for throwing-elbow injury risk and performance.