Shoulder Strength and Physical Activity Predictors of Shoulder Pain in People With Paraplegia From Spinal Injury: Prospective Cohort Study
- Sara J. Mulroy,
- Patricia Hatchett,
- Valerie J. Eberly,
- Lisa Lighthall Haubert,
- Sandy Conners and
- Philip S. Requejo
- S.J. Mulroy, PT, PhD, Pathokinesiology Laboratory, Rancho Los Amigos National Rehabilitation Center, 7601 E Imperial Hwy, Bldg 800, Downey, CA 90242 (USA).
- P. Hatchett, DPT, NCS, KEMG, Pathokinesiology Laboratory, Rancho Los Amigos National Rehabilitation Center.
- V.J. Eberly, PT, NCS, KEMG, Pathokinesiology Laboratory, Rancho Los Amigos National Rehabilitation Center.
- L. Lighthall Haubert, MPT, KEMG, Pathokinesiology Laboratory, Rancho Los Amigos National Rehabilitation Center.
- S. Conners, PTA, Pathokinesiology Laboratory, Rancho Los Amigos National Rehabilitation Center.
- P.S. Requejo, PhD, Pathokinesiology Laboratory, Rancho Los Amigos National Rehabilitation Center.
- Address all correspondence to Dr Mulroy at: smulroy{at}dhs.lacounty.gov.
Abstract
Background Shoulder joint pain is a frequent secondary complaint for people following spinal cord injury (SCI).
Objective The purpose of this study was to determine predictors of shoulder joint pain in people with paraplegia.
Methods/Design A 3-year longitudinal study was conducted. Participants were people with paraplegia who used a manual wheelchair for at least 50% of their mobility and were asymptomatic for shoulder pain at study entry. Participants were classified as having developed shoulder pain if they experienced an increase of ≥10 points on the Wheelchair User's Shoulder Pain Index in the 3-year follow-up period. Measurements of maximal isometric shoulder torques were collected at study entry (baseline), 18 months, and 3 years. Daily activity was measured using a wheelchair odometer, and self-reported daily transfer and raise frequency data were collected by telephone every 6 weeks.
Results Two hundred twenty-three participants were enrolled in the study; 39.8% developed shoulder pain over the 3-year follow-up period. Demographic variables and higher activity levels were not associated with shoulder pain onset. Baseline maximal isometric torque (normalized by body weight) in all shoulder muscle groups was 10% to 15% lower in participants who developed shoulder pain compared with those who remained pain-free. Lower shoulder adduction torque was a significant predictor of shoulder pain development (log-likelihood test=11.38), but the model explained only 7.5% of shoulder pain onset and consequently is of limited clinical utility.
Limitations Time since SCI varied widely among participants, and transfer and raise activity was measured by participant recall.
Conclusions Participants who developed shoulder pain had decreased muscle strength, particularly in the shoulder adductors, and lower levels of physical activity prior to the onset of shoulder pain. Neither factor was a strong predictor of shoulder pain onset.
Shoulder joint pain is one of the most frequent secondary complaints of people following spinal cord injury (SCI), with a prevalence reported to range from 36% to 71%.1–5 Shoulder pain in the SCI population is most commonly related to rotator cuff impingement and tears6,7 and has been attributed to the increase in upper extremity (UE) weight bearing following lower extremity paralysis.4,8,9 Because individuals with SCI are reliant on their arms for mobility as well as daily care activities, shoulder pain negatively affects their quality of life and social participation.10
Although the increased prevalence and negative impact of shoulder pain on quality of life after SCI have been documented for several decades, it remains a common and often untreated problem.7,11 Significant amelioration of shoulder pain has been documented in people with SCI who underwent stretching and strengthening exercise interventions, but pain relief was not complete for all participants.12,13 Moreover, surgical repair, commonly used to treat rotator cuff tears in the nondisabled population, is often not successful for individuals with SCI because of the extended rest period of the affected arm required after surgery.14,15 Thus, the negative impact on quality of life and difficulty treating shoulder pain highlights the need for a preventative program to preserve shoulder function for people after SCI.
To design an effective program to prevent shoulder pain for people with SCI, it is necessary to identify variables that contribute to its increased prevalence and develop interventions for those risk factors that are amenable to change. Although demographic variables such as age,11 sex,16 duration of SCI,1,17 and level of injury18 have been associated with a higher prevalence of shoulder pain onset in some studies, they cannot be altered. As shoulder pain after SCI is predominantly an overuse phenomenon for those with paraplegia and intact innervation of UE musculature, we wanted to evaluate factors that increase the shoulder demand (load and repetition) as well as the person's capacity to meet those demands (muscle strength).
Transfers, depression pressure-relief raises, and wheelchair propulsion (WCP) are the UE weight-bearing activities commonly performed by people with SCI that have been linked to increased shoulder pain.8,9 Transfers and raises are performed multiple times throughout the day and involve high shoulder forces to lift body weight on fixed arms.19,20 Wheelchair propulsion creates relatively low to moderate forces at the shoulder,21 but the frequency of propulsion cycles per day is high (1.9–2.5 km per day, equating to several thousand individual pushes).22–24 In an animal model of rotator cuff pathology, both extrinsic compression and high repetition overuse were shown to contribute independently to supraspinatus tendinosis with additive negative effects when both factors were present. These results implicate both high-force and high-repetition activities as potential contributors to rotator cuff pathology.25
In cross-sectional studies, people with SCI who reported shoulder pain perceived pain-related limitations in WCP, self-care, sports, and leisure activities, but no differences were found in reported frequency of transfers, participation in sports, employment status, or driving between those with and without shoulder pain.8,11,17 These findings imply that people with shoulder pain continued to perform activities of daily living despite having pain. Greater limitations have been found in the frequency of participation in optional social activities in people who have shoulder pain.10,26 In contrast, participants in more rigorous, elite-level wheelchair sports were actually less likely to report shoulder pain than nonathletic wheelchair users.2 A prospective research design is needed to identify activity patterns prior to the onset of shoulder pain and determine if any specific activities are related to the development of pathology.
The capacity to meet the demands of UE weight-bearing activities after SCI is related to strength of the shoulder girdle muscles.27 The demands of UE weight-bearing activities are high but not evenly distributed across all muscle groups. During ramp WCP, the ratio of the net joint moment (external demand) to maximal isometric torque (muscle strength capacity) was greatest for shoulder flexion (67%), followed by external rotation (34%).27 During the body lift portion of sitting pivot transfers, electromyographic activity as a percentage of maximal effort contraction was highest in the sternal pectoralis major muscle of the leading arm (81%) and moderate in the rotator cuff (infraspinatus muscle 44% and supraspinatus muscle 38%), indicating high and moderate demands in these muscle groups, respectively.28
Reduced strength has been documented in those shoulder muscle groups with high functional demands in people with SCI who exhibited shoulder impingement syndrome compared with those with SCI and asymptomatic shoulders.18,29 These investigators posited that reduced strength in the humeral head depressors (lower rotator cuff and adductors) contributed to development of rotator cuff pathology by providing inadequate depression of the humeral head during weight-bearing activities.18,29 What is unclear is whether the reductions in muscle strength were present prior to the shoulder pathology and contributed to its onset or whether the decreased strength was simply a response to the pain and limited function. A prospective study of shoulder pain development over the first year after SCI identified that lower muscle strength (summed UE manual muscle tests) was predictive of shoulder pain onset.30 That study did not identify the impact of strength in individual muscle groups and did not evaluate predictors of longer-term shoulder pain.
Our long-term goal is to develop guidelines for less stressful shoulder function and prevention of chronic shoulder pain for people after SCI. The purpose of this prospective study was twofold. First, we wanted to determine if daily activity levels of the primary UE weight-bearing functions increase the risk of shoulder pain development over a 3-year time period for individuals with paraplegia from SCI who use a manual wheelchair for mobility. Second, we wanted to identify if greater strength, as measured by maximal isometric torques in the key shoulder muscle groups, is associated with protection from development of shoulder pain. We hypothesized that higher frequency of transfers and raises and greater WCP activity (daily distance and average speed) would be significant risk factors for development of shoulder joint pain and that increased shoulder strength (maximal isometric torques) for a given body weight would reduce the risk of developing shoulder pain. Furthermore, we hypothesized that the torques generated by the lower rotator cuff muscles (internal and external rotation) and the primary thoracohumeral depressor muscles (adduction and extension) would be the strength values most related to protection from shoulder pain.
Method
Design Overview
All participants were evaluated at study entry (baseline), including a clinical evaluation of the shoulder, measurement of body weight and maximal isometric torque for all major muscle groups at the shoulder, documentation of wheelchair sports participation, and recording of the biomechanics of both upper extremities during free, fast, and inclined (graded) WCP. An odometer was installed on the wheelchair of each participant at the baseline evaluation, and the cumulative averages of daily distance and speed of WCP in the home and community were documented every 6 weeks during the 3-year study period. Participants also were asked to recall the frequency of transfer and raise activity in the 24 hours preceding each telephone call. Participants returned at 18 months and 3 years following the initial evaluation. Presence of shoulder pain, isometric torques, and WCP biomechanics were reassessed. The WCP biomechanics results are not reported in this article, which focuses on the shoulder muscle strength and home and community activity measures. The presence or absence of shoulder pain over the 3-year study was used to stratify participants into pain and no-pain groups. Shoulder muscle strength (maximal isometric torques) at baseline, daily WCP activity levels (odometer readings), and self-reported transfer, raise, and hours per week spent in wheelchair sports activities were assessed as predictors of shoulder pain development.
Setting and participants.
Participants were self-selected and volunteered in response to flyers posted at outpatient clinics at Rancho Los Amigos National Rehabilitation Center (RLANRC). Informed consent was obtained, and prospective participants were screened to determine eligibility. People were candidates for inclusion into the study if they: (1) had complete or incomplete paraplegia from SCI with a neurological level of injury according to the International Standards for the Neurological Classification of Spinal Cord Injury (ISNCSCI)31,32 of T2 or lower, (2) had a duration of SCI from 2 to 20 years, (3) were ≥18 years of age at study entry, (4) had an absence of shoulder pain with a total score of ≤12 on the Wheelchair User's Shoulder Pain Index (WUSPI),33 (5) used a manual wheelchair for mobility at least 50% of the time, and (6) were able to understand informed consent. They were not admitted into the study if any of the following exclusion criteria were present: (1) history of shoulder fracture or surgery; (2) presence of rotator cuff tendinopathy, bicipital tendinitis, adhesive capsulitis, or cervical radiculopathy; (3) positive findings on any clinical tests for rotator cuff impingement or tear (Jobe Empty Can Test,34,35 Codman Drop Arm Test,36,37 Hawkins-Kennedy Impingement Test,37 and resisted external rotation,37); or (4) any serious medical conditions. Prior to data collection, participants read and signed an informed consent statement that had been approved by the RLANRC Institutional Review Board.
Equipment and procedures.
Shoulder pain at study entry was ruled out using the WUSPI.33 The WUSPI is a 15-item questionnaire that asks a person to mark his or her level of shoulder pain within the previous 7 days from 0 (“no pain”) to 10 (“worst pain ever experienced”) on a 10-cm visual analog scale during different activities, including transfers (4), wheelchair mobility–prolonged and incline (2), self-care activities (5), and general activities (4).33 The WUSPI has been documented to be highly reliable with repeated administration (r=.9).38 To correct for items not performed by the individual, the sum of responses on all performed items is divided by the number of activities performed and multiplied by 15 for a total WUSPI score. The criterion for entry into this study was a total WUSPI score of 12 or less (average of 0.75 per item). We used a threshold for inclusion in the group that developed shoulder pain over the 3-year study of an increase in the total pain score of 10 points or greater from the total score in the initial assessment in either the 18-month or 3-year follow-up evaluations. These values were selected a priori and based on our clinical experience with the questionnaire and on a study of wheelchair athletes with and without shoulder pain.39 Wheelchair athletes with pain had a mean total WUSPI score of 26.3 (SD=22.6), and the asymptomatic group had a mean score of 3.2 (SD=6.2).
Body weight was measured using a platform scale (Arlyn Scales, New York, New York, http://www.arlynscales.com) that the participant rolled onto in his or her wheelchair. The participant then transferred onto a table, and the wheelchair was weighed without the participant. Body weight was calculated by subtracting wheelchair weight from the total weight. Participants were then asked whether they participated in wheelchair sports activities and to estimate the average number of hours in a typical week they spent engaging in the sports activities.
Peak maximal isometric torques were measured using a Biodex System 3 Pro dynamometer (Biodex Medical Systems Inc, Shirley, New York) to assess bilateral shoulder strength for a total of 6 muscle groups: shoulder flexors, extensors, abductors, adductors, internal rotators, and external rotators. The participant was tested in a seated position with the trunk and pelvis secured using 2 chest straps and a lap strap. Shoulder flexion and extension torques were measured with the UE positioned in 45 degrees of flexion and with the forearm in neutral. Shoulder abduction and adduction torques were assessed with the UE positioned in 45 degrees of abduction and with the forearm in neutral. External and internal rotation torques were evaluated with the UE positioned in 90 degrees of abduction, the elbow in 90 degrees of flexion, and the forearm in neutral. Each participant was instructed to push or pull against the lever arm using maximum voluntary effort for 5 seconds. Two trials were performed, with a 15-second rest break between trials.
Daily home and community WCP was measured using the ToPeak bicycle odometer system (ToPeak Inc, Taichung, Taiwan). Each participant's wheelchair was equipped with a sensor mounted onto the wheelchair frame, 2 magnets mounted to each wheel (so the wheels could be interchangeable), and an odometer containing a computer chip to capture the data (Fig. 1). The odometer was calibrated according to the wheelchair wheel circumference. Data were collected on all manual wheelchairs used more than 4 hours per week. Therefore, participants who used an additional wheelchair, such as for sports, had the second wheelchair similarly equipped. Participants were contacted by telephone approximately once every 6 weeks over the 3-year study period to collect and record odometer information, including cumulative averages of daily distance traveled and velocity of propulsion over the 6-week period since the previous call. Telephone calls were made only on weekdays. During these telephone calls, participants were asked to recall the number of car transfers, other types of transfers, and depression pressure-relief raises performed over the previous 24-hour period. We assessed frequency of car transfers separate from other transfers because this activity has been documented to provoke high levels of shoulder pain11 and create the highest physical strain measured by increase in heart rate compared with other types of transfers, WCP, and various activities of daily living.40
Example of the wheelchair odometer setup. Each wheelchair was equipped with a sensor (circled), 2 magnets (arrows), and an odometer containing a computer chip (rectangle).
Data Analysis
Group assignment-onset of shoulder pain.
Participants who had an increase in the total WUSPI score of 10 points or greater from the initial evaluation were classified as having developed shoulder pain, and those with a smaller increase, no change, or a decrease were classified as asymptomatic. Participants who reported shoulder pain were asked if their pain was in the right, left, or both shoulders.
Daily activity and shoulder torque predictor variables.
The mean for each of the activity variables (WCP distance and average speed and number of car transfers, other transfers, and depression pressure-relief raises performed in 24 hours) was calculated for all readings taken prior to the onset of shoulder pain for those who developed pain and over the entire 3-year period for those who remained pain-free. The peak isometric torques of the 2 trials for each muscle group were averaged and normalized to body weight for each UE. We elected to normalize torque values by body weight because many mobility activities after SCI involve moving or lifting body weight with arms. In addition, normalizing by body weight allows comparison of shoulder torques between men and women and among people of various sizes. For participants who remained pain-free and for those who developed bilateral pain, the peak isometric torque was averaged for right and left sides for each shoulder muscle group tested at the initial assessment. For participants who developed unilateral shoulder pain, torque values on the side of shoulder pain development were evaluated.
Statistical Analysis
Statistical analyses were conducted at the .05 significance level using SPSS version 12.0 (SPSS Inc, Chicago, Illinois). Step-wise logistic regression was conducted to evaluate demographic parameters as potential covariates (step 1) and then to determine whether daily activity variables (step 2) and maximal isometric shoulder torques (step 3) predicted the development of shoulder joint pain (yes or no). We used forward stepping with an F value to enter of 0.05 and an F value to remove of 0.10. The log-likelihood ratio chi-square value was the criterion used when considering which variables to include in the logistic regression model. We first evaluated age, time since SCI, neurological level of SCI, and sex as potential covariates in the models to predict shoulder pain onset.
We assessed the potential independent variables for multicollinearity, which can inflate variability of predictor coefficients and make the model unstable, first by examining the bivariate correlations and then by evaluating the variance inflation ratios of predictors in the final model.41 For the activity variables, the bivariate correlations were low to moderate (.01–.33), indicating low levels of multicollinearity. Bivariate correlations of the muscle torque variables, however, were higher (.57–.84). Correlations were particularly high between pairs of torque values from muscle groups with similar functional demands, including the flexors and abductors (both humeral elevators, r=.81) and the adductors and extensors (both thoracohumeral depressors, r=.84). To reduce potential multicollinearity, we selected a representative muscle from each of the 3 functional groups—humeral elevators, thoracohumeral depressors, and rotator cuff—to include as potential predictors in our model. We selected the representative muscle groups based on both the bivariate correlations with shoulder pain onset and their magnitude of activity in daily functions, including transfers, depression raises, and WCP.27,42,43 The 3 torque values selected were flexion, adduction, and external rotation. Variance inflation factors and odds ratios for each independent variable included in the final equations were calculated, and the likelihood ratio and Hosmer and Lemeshow chi-square tests for goodness of fit were performed.
Role of the Funding Source
This study was supported by a grant from the National Institutes of Health (grant #R01 HD049774).
Results
Participant Recruitment, Retention, and Demographics
Eligible participants were recruited from outpatient clinics at RLANRC from October 2006 to March 2010. Among the 327 individuals screened for eligibility, a total of 223 participants (198 men and 25 women) were eligible, agreed to participate, and signed an informed consent statement (Fig. 2). The average age of participants at baseline was 34.7 years (SD=9.3), mean time since SCI was 9.3 years (SD=6.2), and average body weight was 74.1 kg (SD=16.3). The majority of the participants (n=129) had lower-level paraplegia (T8 and below), and 94 participants had high-level paraplegia (T2–T7). Follow-up testing at 18 months after the baseline assessment was completed by 209 participants, and 201 returned for follow-up testing at 3 years. Of the 22 participants who did not complete follow-up testing, 15 were lost contacts, 5 were deceased, 1 was incarcerated, and 1 was removed because of prior shoulder surgery that was not identified at enrollment.
Flow diagram of the study.
Participant characteristics at study entry were compared between the 22 participants who did not complete the follow-up assessments and the 201 participants with complete follow-up data using an independent t test. Only age demonstrated a statistically significant difference between groups. Those participants who were lost to follow-up were younger (mean age=30.2 years, SD=6.7) compared with those who completed the study (mean age=35.0 years, SD=9.4) (P<.022). Three participants were lost to follow-up after the initial evaluation but returned for the 18-month and 3-year follow-up assessments, so no community WCP or transfer and raise activity levels were recorded. An additional 6 participants declined to have the odometer attached to their wheelchair. Consequently, WCP readings were analyzed for 192 participants, and transfer and raise frequencies were documented in 198 individuals. There were no differences in baseline characteristics between the 9 individuals who did not record their WCP and transfer data and the 192 participants with complete data.
Group Assignment—Onset of Shoulder Pain
A total of 80 participants (39.8%) (out of the 201 who completed follow-up testing) developed shoulder pain at some point over the 3-year study period. Of those who developed shoulder pain, 18 had pain in the left shoulder only, 32 had right shoulder pain only, and 30 had pain in both shoulders. Shoulder pain had developed at the 18-month follow-up assessment in 55 participants (27.4%), and 25 additional people (12.4%) developed shoulder pain by the 3-year follow-up testing (Tab. 1).
Shoulder Pain Levels (WUSPI Scores) at Baseline, 18-Month, and 3-Year Assessments by Pain Groupa
Demographic Covariates
None of the demographic variables was a significant predictor of shoulder pain onset. Therefore, no covariates were included in the activity and muscle torque predictor model (Tab. 2).
Demographic Variables for the No-Pain and Pain Groupsa
Daily Activity Predictors
None of the WCP (average speed and distance) or other daily activity variables (frequency of car transfers, other transfers, raises, or weekly hours of sports participation) were a significant predictor of shoulder pain onset after controlling for time since SCI. Activity levels were either lower in participants who developed shoulder pain than in those who remained pain-free or were similar in the 2 groups (Tab. 3). Average speed of propulsion was slower in participants who developed shoulder pain (2.4 km/h, SD=0.9) than in those who remained pain-free (2.8 km/h, SD=1.5). Participants who eventually developed shoulder pain performed fewer car transfers per day prior to pain onset (X̅=4.0, SD=3.3) compared with those who remained pain-free (X̅=5.1, SD=3.5). Daily distance of WCP, frequency of depression pressure-relief raises performed each day, and hours per week spent in wheelchair sports were similar in the 2 groups (Tab. 3).
Daily Activity Variables for No-Pain and Pain Groupsa
Shoulder Torque Predictors
Baseline maximal isometric torque (normalized by body weight) in all of the major shoulder muscle groups was 10% to 15% lower in participants who eventually developed shoulder pain than in those who remained pain-free (Tab. 4). Shoulder adduction torque entered the step-wise logistic regression model predicting shoulder pain onset (P=.001) (eTabs. 1 and 2). Once shoulder adduction torque was entered into the model, none of the remaining torque variables added significantly to the prediction of shoulder pain onset.
Peak Isometric Shoulder Torque Variables for No-Pain and Pain Groupsa
Discussion
Using a prospective, longitudinal design, we identified that individuals with paraplegia who developed shoulder pain had weaker shoulder muscle strength (particularly in the adductors) prior to the onset of shoulder pain than those who remained pain-free. Shoulder muscle torque, however, explained only 7.5% of the variability in the model and, consequently, is not a clinically useful predictor of shoulder pain onset. Contrary to our hypothesis, higher UE weight-bearing activities were not risk factors for shoulder pain development. Those participants who eventually developed shoulder pain were less active prior to the onset of shoulder pain than those who remained pain-free. They performed fewer car transfers and propelled at a slower velocity on a daily basis than those who remained pain-free. Lower activity levels, however, were not a significant predictor of shoulder pain onset. These results suggest that, for some individuals who eventually developed shoulder pain, even the minimum requisite daily demands of UE weight bearing after SCI created an increased risk for shoulder pain development over time.
Isometric torque in all of the shoulder muscle groups tested was lower in participants who eventually developed shoulder pain, particularly shoulder adduction torque. This finding likely relates to the high demand on the adductors during common weight-bearing functional activities for people with paraplegia. Shoulder adduction and flexion moments are high during depression transfers and moderate during WCP and depression raises.20,21,27,44 The pectoralis major muscle, the primary adductor of the shoulder, is also a shoulder flexor, particularly when the shoulder is extended or in low angles of elevation when both the clavicular and sternal portions have a flexion moment arm.45 This shoulder adductor muscle is intensely active during transfers as well as the push phase of WCP and moderately active in depression raises.28,42,43,46,47 Owing to its attachments on the proximal humerus and more distally on the thorax, during UE weight-bearing activities such as transfers and raises, the sternal portion of pectoralis major muscle lifts the trunk (and scapula) off of the humeral head and unloads the rotator cuff.28 People with weaker shoulder adductors may have a less effective unweighting mechanism, particularly with fatigue.
In contrast, increased strength in the shoulder external rotator muscle group was the strongest mediating factor in reduction of shoulder pain in people with SCI who performed a 12-week shoulder home exercise program described in articles on the Strengthening and Optimal Movements for Painful Shoulders (STOMPS) study.12,26 It is possible that inadequate adductor strength contributes to the development of the shoulder pathology (ie, impingement), but once pain is present, strengthening of the external rotators is more critical to pain reduction. Participants in the STOMPS study had a longer duration and greater intensity of shoulder pain than those who developed pain over the 3-year period in this study (mean WUSPI score of 48 in the STOMPS study versus 33 in this study).12 Maximal torques for adduction and internal rotation were similar for the participants who eventually developed shoulder pain in this study and those in the STOMPS study prior to the exercise intervention. In contrast, abduction and external rotation torques were approximately 17% lower in the STOMPS study pre-exercise values than in our pain group. These differences could be related to varying methods and positions of strength testing between the 2 studies, or they could represent a progression in rotator cuff weakness with a longer duration of pain. This scenario would indicate that untreated shoulder pain increases in intensity over time and external rotation and abduction strength continues to decline.
The magnitude of differences in shoulder torques between the pain and no-pain groups in this study was moderate at 10% to 15%. This magnitude of strength increase is similar to that seen in people with SCI who participated in 12-week home exercise or circuit training programs, indicating that it is a feasible goal with an exercise program.12,48 Van Drongelen and colleagues30 identified a similar magnitude for the relationship between UE muscle strength and development of shoulder pain in the early post-SCI period for a cohort of participants with paraplegia or tetraplegia. They found that a 17% increase in the summed UE manual muscle test score would predict a 12% reduction in shoulder pain at 1 year after discharge from inpatient rehabilitation. Given the lack of fit of the predictor model in the current study, it is not clear that a shoulder muscle strengthening in the chronic phase after SCI would significantly reduce the risk of shoulder pain development.
Reduced shoulder muscle strength is likely not the only or the major factor contributing to development of shoulder pain after SCI, as our predictive model explained only 7.5% of the variation in shoulder pain onset (Nagelkerke R2=.075, eTab. 2) and the goodness-of-fit test of the model did not reach statistical significance (P=.15). Therefore, we recommend shoulder muscle strengthening primarily for improved functional capacity. Strengthening exercises for the major shoulder muscles in people with SCI result in increased capacity for muscle work and the high-intensity activities of daily living48,49 and reduce shoulder pain when present.12,13,26,50 Shoulder muscle strengthening for prevention of shoulder pain after SCI was not strongly supported by the results of this study. It is possible that, over time, the added demands of the various UE mobility activities required after SCI exceed the muscle capacity of most people. Despite the small and clinically questionable reduction in risk of shoulder pain development seen with stronger shoulder adductors in our cohort of participants in the chronic post-SCI phase, and because few alternative strategies are available for prevention of shoulder pathology in this population without a significant alteration in mobility such as using a power wheelchair and transfer aids, we still recommend shoulder strengthening exercise for the functional benefits.51
The findings of this study shed some light as to why information in the literature is conflicting as to whether wheelchair sports are protective or a risk factor for shoulder pain development.2,17,39 Wheelchair sports can increase daily propulsion activity. Sporner and colleagues52 documented that wheelchair athletes propelled similar distances at almost twice the speed during 1 hour of wheelchair basketball and rugby as our participants covered in a full day of community propulsion. This finding represents a significant increase in daily demand on the shoulder. Similar to the findings of Finley and Rodgers,17 we found that wheelchair sports participation by itself did not affect the risk of shoulder pain development. In contrast, Fullerton and colleagues2 documented that elite wheelchair athletes were 2 times less likely to report shoulder pain than a nonathletic comparison group, and those athletes who did report shoulder pain developed it 4 years later (after SCI onset) than nonathletes. The investigators did not report shoulder muscle strength or body weight, but it is possible that the elite athletes also engaged in shoulder muscle strengthening, which could have been protective against the onset of shoulder pain. Finley and Rodgers17 reported that participants in their study engaged in a range of recreational to elite levels of competition, whereas wheelchair athletes in the current study were primarily recreational athletes, and few reported engaging in muscle strengthening exercises. Engagement in wheelchair sports has numerous physiological and psychological benefits,53–56 but wheelchair athletes should be encouraged to include a muscle strength training regimen to ensure the capacity to tolerate the increased shoulder demands of sports activities.
Limitations
Our study had several limitations. The primary limitation was the design of participant recruitment. To limit bias in a cohort study, participants should be representative of the total population and enter the study at a common time point in the course of their disease. In this study, all participants were pain-free at study entry, but the time since SCI varied widely (2–20 years). Development of shoulder pain in the first few years after SCI mostly occurs in people with cervical-level injuries resulting in tetraplegia and is related to spasticity and reduced passive shoulder joint mobility.3,57 In a cross-sectional study, the prevalence of shoulder pain in people with paraplegia was found to be relatively low for the first 5 years after SCI (18%) and to increase from 5 to 20 years postinjury, affecting more than 70% of those with a duration of SCI greater than 20 years.3 The logistics of following participants prospectively from onset of SCI for 10 to 20 years are prohibitive. Consequently, we enrolled participants with varying durations of SCI who had not yet developed shoulder pain. Individuals who had already developed shoulder pain were excluded from this study, creating a potential selection bias in those with longer duration of SCI. Also, participants who were lost to follow-up were approximately 5 years younger than those who completed the study. As age was not related to shoulder pain onset, we do not believe the omission of the younger participants affected our results.
An additional limitation of our study was that we did not directly measure the number of transfers and raises performed by our participants but asked them to recall the frequencies over the previous 24-hour period. Our participants reported an average of 12 total transfers (both into and out of the wheelchair) and 12 depression raises per day. This rate is higher than that found by Yang and colleagues,58 who measured wheelchair seat pressures and recorded an average of 9.4 “lift-off” activities per day (including both raises and transfers out of the wheelchair). Even considering that we counted transfers both into and out of the wheelchair, our participants reported significantly more raises and transfers than measured by Yang and colleagues. Part of this discrepancy could be due to recall error.
Pressure-relief raises occur throughout the day and are not tied to other activities, likely making them harder to accurately recall than transfers. Car transfers occur less frequently, are associated with trips into the community, and thus may be easier to recall accurately than other transfers and raises. Our participants reported performing, on average, between 4 and 5 car transfers per day. A single car trip to and from one location would require 4 car transfers (into the car initially, out and back into the car at the destination, and back out of the car at home). As many of our participants did not drive or ride in a car out of their wheelchair and some likely took trips to more than one destination per day, the mean number of car transfers reported seems reasonable. Weekly hours of wheelchair sports participation was estimated only at study entry, but this activity also would have been captured in WCP odometer readings.
Only shoulder muscle strength was evaluated in this study. Elbow flexor and extensor strength, in particular, also are critical to optimal performance of transfers, raises, and WCP.28,46,47,59 Finally, our results are only applicable to individuals with paraplegia. Individuals with tetraplegia typically have earlier, more frequent, and more severe shoulder pain than those with paraplegia.3,30 It is likely that the relationships among muscle strength, activity level, and risk of shoulder pain development differ between people with tetraplegia who have impaired innervation in the UE muscles and those with paraplegia.
In conclusion, individuals with paraplegia from SCI who were asymptomatic at study entry but developed shoulder pain over a 3-year period had weaker maximal isometric shoulder torques than those who remained pain-free. The overall reduction in risk of shoulder pain development from shoulder muscle weakness was small, however, and was not a clinically useful predictor. Daily physical activity levels were not related to risk of shoulder pain development, and those participants who developed pain were actually less active even prior to the onset of shoulder pain. A prevention program of shoulder muscle strengthening for people with paraplegia is recommended to improve or maintain capacity for function even though it may not reduce the risk of shoulder pain development.
Footnotes
Dr Mulroy and Dr Requejo provided concept/idea/research design. Dr Mulroy, Ms Hatchett, and Dr Requejo provided writing. Ms Hatchett, Ms Eberly, Ms Lighthall Haubert, Ms Conners, and Dr Requejo provided data collection. Dr Mulroy, Ms Hatchett, Dr Requejo, and Ms Conners provided data analysis. Dr Mulroy and Dr Requejo provided project management and fund procurement. Dr Requejo provided facilities/equipment. Ms Conners provided administrative support. Ms Hatchett, Ms Eberly, Ms Lighthall Haubert, and Dr Requejo provided consultation (including review of manuscript before submission). The authors acknowledge the technical support of Ramon Cervantes, Juan Garibay, Diego Rodriguez, Maneekobkunwong, Somboon, Charles Whitehead, and Carlos Williams.
This study was approved by the Institutional Review Board of Rancho Los Amigos National Rehabilitation Center.
This study was supported by a grant from the National Institutes of Health (R01 HD049774).
- Received December 17, 2013.
- Accepted February 17, 2015.
- © 2015 American Physical Therapy Association