Prevalence of Upper Limb Pain in Spinal Cord Injury: A Systematic Review

Abstract

A systematic review was undertaken to evaluate and critically appraise literature pertaining to prevalence and treatment of upper limb pain in the spinal cord injured (SCI) population using manual wheelchair. Data extraction tables were compiled, then an in-depth data on the types of injury, level of injury, type of wheelchair used, type of treatment sought and the impact on Activities of Daily Living were recorded. A Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies tool was used to critically appraise the quality of studies included in this review. 994 papers in total were screened, 46 full text studies were assessed with 14 studies included in the final synthesis: four cohort studies and ten cross-sectional studies. Shoulder pain was the most common type of pain reported (30–71%) followed by wrist, hand, and elbow. Functional limitations reported because of upper limb pain included interference with mobilizing, transferring, and Activities of Daily Living, primarily personal care tasks. There is clear evidence that upper limb pain is prevalent in the SCI manual wheelchair using population which impacts on functional tasks. Further research is required to explore the perceptions of those with upper limb pain and techniques used to manage pain.

1. Introduction

Participant with Spinal cord injury (SCI) have been reported to experience premature or accelerated aging in several organ systems in the SCI population compared to the aged matched general population [1]. In addition, they report that chronic pain and other health conditions increases with the duration of SCI. The primary complications that can occur in the short and long term after SCI include musculoskeletal (MSK) pain, muscle atrophy, pressure sores, infections, and respiratory issues [2].

The scope of this review is in relation to Musculoskeletal (MSK) pain, specifically of the upper limb (Figure 1). For the purpose of this study, upper limb pain refers to pain or inflammation of the neck, shoulder, elbow, wrist, or fingers as well as the corresponding muscles, ligaments and tendons. There is a substantial amount of literature in the area documenting the prevalence of these conditions. Injuries such as shoulder, neck and back pain resulting from poor wheeling practice in the long-term are documented in both those who began wheeling as adults and as children [3, 4, 5]. Between 49% and 73% of SCI manual wheelchair users develop carpal tunnel syndrome and between 31% and 71% report shoulder pain [6]. This may have serious implications for functional mobility, sleep and living life independently [7].

Figure 1.

Prevalence of upper limb pain in the spinal cord injured (SCI) population using manual wheelchair. Adapted from: incountryvalueoman.net. Reproduced with the kind permission of Leda Bug (@art_and_spacecrafts). 1: Shoulder; 2: Wrist; 3: Hand; 4: Elbow.

Management of upper limb pain may prove difficult due to the nature of the treatment. In many cases relative rest may be required for the upper limb to recover however this may prove problematic as the upper extremity is used for mobility on a daily basis [8]. Pain can contribute to overall poorer health in the SCI population and has been shown to have a negative effect on both physical and psychological aspects of a person’s wellbeing [9]. Further long-term pain that is chronic in nature has also been associated with low mood and depressive symptoms in the SCI population [10].

2. Aim

The overall aim of this review is to examine the literature in relation to prevalence of upper limb pain, pain sites reported, treatments availed of and causation of injuries. This review follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [11].

3. Methods

3.1 Search and study selection

A search was conducted between January–February 2019 for studies reporting on the prevalence of upper limb injuries or pain, in manual wheelchair users with an SCI. Medline (1966 – February 2019), CINAHL (1982 – February 2019), OVID (1966 – February 2019) and PubMed (1971 to February 2019) databases were searched using the terms “spinal cord injur* or SCI” combined with “wrist”, “elbow”, “shoulder”, “neck”, “upper limb”, “carpal tunnel”, “rotator cuff”, “parapleg*”, and “mobil*”, “ambulation”, “propel”, and “pain”. Further literature was obtained by exploring reference lists of papers identified in this search. Each title was screened by a single reviewer for relevance and added to the shortlist if it met the inclusion criteria or if further clarification was required, the abstract or entire paper was reviewed.

3.2 Inclusion and exclusion criteria

Studies were included if they were peer reviewed research studies written in the English language, that directly reported on prevalence of upper limb pain in SCI. Studies were required to include participants with a traumatic SCI only and use a manual wheelchair full time. Other causations of SCI were excluded such as infection or insufficient blood flow, as in these cases participants may regain function and therefore fluctuating prevalence rates of upper limb pain may be observed. Any prevalence rates reported in these studies may be skewed by a participant regaining function or not requiring a wheelchair for mobility purpose therefore would not be an accurate reflection of the true prevalence rates. Studies primarily including wheelchair athletes were also excluded as it is common for athletes to have higher level of activities compared to a sedentary population and may therefore report higher levels of prevalence rates that could not be generalized to the wider SCI population.

3.3 Data collection process

Data extraction tables were compiled (Appendix 1) and included study design, objective, sample size, classification of SCI, type of injury/pain reported, outcome measures used and results of each article. Further in-depth data on the types of injury recorded, level of SCI, type of wheelchair used, type of treatment sought (if applicable) and the impact on ADLs, were also recorded.

3.4 Study quality appraisal

The National Heart, Lung, and Blood Institute (NHLBI) Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies tool was used to critically appraise the quality of studies included in this review. The tool is a widely accepted tool used for appraising observational studies and is particularly useful in identifying methods applied to minimize bias in research literature [12]. The tool itself is a 14-item scale, with each question scored as “yes” or “no”. If an item on the checklist cannot be clearly identified, the scorer can assign “cannot determine”, “not applicable” or “not relevant”. The tool has been designed as a checklist rather than a scoring scale specifically, however, can be used as guidance in determining the methodological quality of studies. All studies were retrieved and reviewed by a single researcher (AMC).

4. Results

The systematic search returned 994 papers in total (Figure 2: PRISMA Flow Diagram). Two additional papers were found via hand search and review of relevant reference lists in the subject area. Forty-six studies were selected for further reading. After reviewing the full text studies, 31 studies were excluded after not meeting one or more of the inclusion criteria. The most common inclusion criteria not met was the involvement of part time manual wheelchair users, elite wheelchair athletes or studies not specific to upper limb or extremity injury in the SCI population. The total number of studies included in this review was 15 papers.

Figure 2.

PRISMA flow diagram.

Key results for all studies are summarized in Appendix 1. Four studies comprised of cohort methodologies [3, 13, 14, 15] following patients for 3 years, 5 years, 18 months and 1-year post SCI rehabilitation respectively. The remaining 11 studies were cross-sectional in design [16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26]. Five studies investigated the prevalence of upper limb pain alone and the remaining 10 studies, included the impact on functional activities.

4.1 Demographic results

Demographic details from each study are outlined in the appendix (Appendix 1). Studies are discussed in further detail below.

4.2 Recruitment

Studies were primarily conducted in the United States of America (USA), (n = 9), two studies conducted in the Netherlands, two in Australia, one in Sweden and one in Israel. Recruitment of participants was primarily conducted via hospital discharge lists (n = 7). Ballinger et al. [13], additionally advertised their study with local radio stations and Boninger et al. [17], advertised with known wheelchair vendors to improve recruitment. 19 Eriks-Hoogland et al. [19], Silfverskiold & Waters [15] and Van Drongelen et al. [3], recruited participants while they were undergoing initial inpatient rehabilitation. Pentland & Twomey [22, 23] stated participants were recruited from the community however it is not clear whether this may have been via discharge lists, advertisements in the media or any other approach. Escobedo et al. [20] and Sie et al. [25] recruited participants directly on attending a routine medical examination at an outpatient appointment as part of their SCI rehabilitation. The remaining studies [16, 18] do not state explicitly where participants were recruited from.

Research study settings refers to where the study took place. Settings were classified as either inpatient, outpatient or community based. Five studies were community based [21, 22, 23, 24, 26]. Four studies were outpatient based [18, 19, 20, 25]. Two studies were inpatient based [3, 15], and three availed of a combination of community and outpatient settings [13, 17, 19]. Recruitment methods and setting were unclear for one study [16].

4.3 Response rates

Response rates were detailed in five studies: 76.5% [18], 86%, [19], 46% [21], 63% [24] and 66% [26]. Ballinger et al. [13], reported an oversubscription to their study; 661 participants responded with the authors choosing a sample of 140 participants. Escobedo et al. [20] and Sie et al. [25] used a sample of convenience from patients attending routine outpatient appointments and therefore all patients who met the inclusion criteria were included. The remaining studies did not list response rates specifically, however Eriks-Hoogland et al. [19] reported 60 patients were lost to follow up: 43% dropout rate at the end of the five- year study. The remaining cohort studies do not list details relating to dropout rates or participant retention.

4.4 Sample sizes

Sample sizes ranged from 11 participants [22] to 669 participants [21] in a cross sectional study. Sample sizes for each individual study are outlined in Appendix.

4.5 Age

The youngest participant in all studies was aged 17 years [15], with the oldest participant aged 78 years [20]. Eight studies included age range and mean, six studies reported mean age and one study to record age range [19]. The breakdown of reporting methods for age are outlined below in Table 1.

Reporting method	Author	Age
Studies reporting age range and mean	Aljure et al. [16]	Range = 20–73 years, mean = 47.8
	Ballinger et al. [13]	Range = 19–73 years, mean = 37
	Eriks-Hoogland et al. [14]	Range = 18–66 years, mean = 34
	Escobedo et al. [20]	Range = 40–78 years, mean = 59
	Gironda et al. [21]	Range = 20–65 years, mean = 50.6
	Sie et al. [25]	Range = 17–71 years, mean = 37.4
	Silfverskiold & Waters [15]	Range = 17–40 years, mean = 25
	Subbarao et al. [26]	Range = 21–77 years, mean = 53
Studies reporting mean age only	Boninger et al. [17]	35 years
	Dalyan et al. [18]	42.2 years ± 12
	Pentland & Twomey [22]	44.3 years
	Pentland & Twomey [23]	42.9 years
	Samuelsson et al. [24]	49 years ± 18
	van Drongelen et al. [3]	59.6 years
Study reporting age range only	El-Essi et al. [19]	Range = 18–59 years

Table 1.

Reporting of age across studies.

4.6 Gender

Two of the older studies did not provide data relating to gender of participants included in their studies [25, 26]. Three studies used a sample composed of male participants only [16, 20, 22]. The remaining studies all reported a higher percentage of male participants compared to female participants as is reflected in the wider population of SCI patients, where males are twice as likely to suffer an SCI compared to females [27]. This is primarily attributed to the fact men are more likely to take part in high-risk activities such as high-speed driving or dangerous sports [28]. The higher percentage of males in this review may also be attributed to the study design of several studies included. Three studies were conducted as part of the Veterans Affair medical centers in the USA. A higher percentage of males enroll in the military in the USA and therefore the potential cohort of participants recruited from may have been male dominated [29]. Pentland & Twomey [22], were the only study to include a female only sample. Apart from this study, the highest percentage of female participants was 32% [17], albeit a small sample size (n = 32).

4.7 Level of injury

The reporting of level of injury varied widely between studies. The terms quadriplegia and tetraplegia both refer to the same classification of injury and are based on the terminology used by individual authors and reflects differences in language used around the world. For the purpose of this study, the term tetraplegia will be used. Six studies referred to participants as either patients with paraplegia or tetraplegia. Four of these studies included participants with paraplegia only [16, 17, 19, 24]. Sie et al. [25] and Silfverskiold & Waters [15] included participants with tetraplegia and paraplegia; 57% tetraplegia, 43% paraplegia and 66.6% tetraplegia and 33.3% paraplegia respectively.

Ballinger et al. [13] and Eriks-Hoogland et al. [14] both reported level of injury using a combination of the terms high/low paraplegia/tetraplegia and the American Spinal Injury Association (ASIA) Impairment Scale (AIS) levels A-D. Ballinger et al. [13] range included; 5% high tetraplegia, 39% low tetraplegia, 45% paraplegia, 11% ASIA class D. 14 Eriks-Hoogland et al. [14] included 34.1% tetraplegia and AIS class A or B. Escobedo et al. [20] and van Drongelen et al. [3] both list level of injury as ranges; T3-L2 and C2-S5 respectively. The remaining five studies also list level of injury as ranges however provide further details on the percentage of participants within each range.

Dalyan et al. [18] provided the most in-depth detail regarding level of injury; C2-C4 = 14.5%, C5-C8 = 35.5%, T1–5 = 7.9%, T6-T10 = 19.7%, T11-L2 = 21.1% and L3-L4 = 1.3%. Gironda et al. [21] grouped participant level of injury into three ranges: T2-T6 = 34%, T7- T12 = 56.1% and L1-L2 = 9.9%. Similarly, Pentland & Twomey [22, 23] used three ranges, however ranges differ by one level of injury within their groups. In 1991 [23], they reported level of injury as; T1-T5 = 9%, T6-T10 = 18% and T11-L3 = 73%. In 1994 [22], injury level was reported as; T2-T5 = 20%, T6-T10 = 40% and T11-L2 = 40%. Finally, Subbarao et al. [26] grouped the reporting of level of injury into four ranges: C1-C4 = 9.2%, C5 – T1 = 34.6%, T12-L1 = 37.9% and L2 and below = 13.1%. The full range of reporting measures for level of injury can be found below in Table 2.

Author (year)	Level of SCI
Aljure et al. [16]	All participants with paraplegia
Boninger et al. [17]
El-Essi et al. [19]
Samuelsson et al. [24]
Sie et al. [25]	57% tetraplegia, 43% paraplegia
Silfverskiold and Waters [15]	66.6% tetraplegia and 33.3% paraplegia
Ballinger et al. [15]	5% high tetraplegia, 39% low tetraplegia, 45% paraplegia,; 11% ASIA class D
Eriks-Hogland et al. [14]	34.1% tetraplegia and AIS class A or B
Escobedo et al. [20]	T3-L2
van Drongelen et al. [3]	C2-S5
Dalyan et al. [18]	C2-C4 = 14.5%, C5-C8 = 35.5%, T1–5 = 7.9%, T6-T10 =; 19.7%, T11-L2 = 21.1% and L3-L4 = 1.3%
Gironda et al. [21]	T2-T6 = 34%, T7-T12 = 56.1% and L1-L2 = 9.9%
Pentland & Twomey [23]	T1-T5 = 9%, T6-T10 = 18% and T11-L3 = 73%
Pentland & Twomey [22]	T2-T5 = 20%, T6-T10 = 40% and T11-L2 = 40%
Subbarao et al. [26]	C1-C4 = 9.2%, C5 – T1 = 34.6%, T12-L1 = 37.9% and L2 and; below = 13.1%

Table 2.

Reporting measures for level of injury.

4.8 Time since injury

Time since injury was reported either as the mean years since injury or the range of years since injury. One study only [13] reported time since injury as the age that SCI occurred; mean = 27 years, range = 14–68 years. Four studies reported time since injury as the mean number of years since injury only: mean = 11.5 years [17], 11.8 years ± 8.5 years [18], 26 years [20], 20.3 years ± 11.1 [21]. Three studies included time since injury as range, 3 months – 42 years [16] and 6–18 months’ post SCI [3, 15]. Four studies included both range and mean time since injury; 1–45 years, mean = 17.4 [22], 5–21 years, mean = 15.2 [23], 1–42 years, mean = 12.1 [25] and 21–77 years, mean = 22.8 [26]. Two studies did not report any details regarding time since injury [14, 19].

4.9 Area of upper limb pain

The most common site of pain investigated was the shoulder alone (n = 7). Of these, Boninger et al. [17] aimed to gain insight into the prevalence of shoulder injuries, however the study was primarily focused on identifying rotator cuff tears in patients with paraplegia. Six studies investigated the prevalence of pain on all the upper extremities; [3, 18, 21, 22, 23, 25] while 26 Subbarao et al. [26] investigated pain at both the shoulder and wrists. Both Aljure et al. [16] and 20 Escobedo et al. [20] were distinctive in that they investigated the occurrence of an injury rather than a pain site alone. Aljure et al. [16] investigated the prevalence of carpal tunnel syndrome (CTS), while Escobedo et al. [20] investigated the prevalence of RCTs in patients with paraplegia.

4.10 Outcome measures

The primary outcome measure used in all studies was a self-reported questionnaire establishing prevalence and location of pain (n = 11). Interviews were utilized in six studies, either by telephone or face to face, to gather demographic data and data relating to prevalence of upper limb pain and injury. Eleven studies also conducted physical exams to establish prevalence and location of pain. Postal questionnaires were utilized in five studies [18, 19, 21, 24, 26]. Of these, Samuelsson et al. [24] and Subbarao et al. [26] used postal questionnaires as an identification method to invite participants to attend a physical exam to further investigate upper limb pain. Nine studies formulated their own questionnaire; four studies used these to collect data relating to prevalence and location of pain [14, 17, 18, 21] and five studies used these to collect demographic data [3, 22, 23, 25, 26]. No standardized outcome measures were used solely to report prevalence of pain.

4.11 Functional outcome measures

The relationship between upper limb pain and functional limitations was formally assessed in eleven studies [3, 13, 14, 15, 16, 17, 18, 19, 20, 21, 26]. Of these, six studies used standardized outcome measures to report functional limitations [3, 13, 14, 19, 21, 24]. An additional two studies formulated their own functional questionnaire based on standardized outcome measures and pilot tested these with steering groups to ensure content and consensual validity was reached [23, 26]. The most commonly used measures were the Functional Impact Measure (FIM) (n = 3), and the Wheelchair User Shoulder Pain Index (WUSPI) (n = 3); both are reliable and valid tools [30, 31]. The FIM is an 18-item questionnaire designed to assess level of disability and patient’s change in health status in response to further disability such as pain or medical intervention. The FIM is a well- documented assessment of functional ability and has been used across a wide range of disability cohorts. In comparison, the WUSPI has been designed specifically for the wheelchair using population, however, is only specific to shoulder pain, not the upper extremity in its entirety.

A wide variety of additional standardized outcome measures were used across all studies including; the Craig Handicap Assessment and Reporting Technique (CHART) (n = 1), the Shoulder Rating Questionnaire (SRQ) (n = 1), the Sickness Impact Profile 68 (SIP68) (n = 1), the Physical Activity Scale for Individuals with Physical Disabilities (PASIPD) (n = 1), the Klein and Bell Activities of Daily Living Scale (n = 1), the Canadian Occupational Performance Measure (COPM) (n = 1), and the Constant Murley Scale (n = 1). Of these, the SRQ and Constant Murley Scale are both specific to shoulder pain, while the remainder are generic tools assessing functional tasks. None of the standardized outcome measures are specifically designed for use with patients with an SCI.

4.12 Physical assessments

Physical assessments of dysfunction were utilized in twelve studies. Of these, eleven studies utilized a standardized method of assessment [3, 13, 14, 15, 16, 17, 20, 22, 23, 25].

Radiographic imaging was utilized in three studies [13, 17, 20]. Radiographic images were taken following clinical protocols for identification of RCTs [13, 17]. Shoulders were x-rayed in anteroposterior position only [13], while Boninger et al. [17] used additional positions of scapular anteroposterior position and supraspinatus position.

Ballinger et al. [13] also conducted a physical assessment of participants using manual muscle testing and range of movement (ROM). ROM was assessed in three additional studies [14, 22, 23]. Eriks-Hoogland et al. [14] assessed physical ROM via manual muscle testing and completion of the Wheelchair Skills Test. Biomechanical measures were taken using peak power output (POpeak) requiring participants to complete a maximal wheelchair exercise test on a motor-driven treadmill.

Transfers were also assessed using the FIM. Pentland & Twomey [22], assessed ROM at both the shoulder and elbow. Bilateral upper limb function was assessed using concentric isokinetic torque using KinCom II isokinetic dynamometer and Smedley’s handheld dynamometer, both which are valid and accurate tools for measuring muscle strength [32, 33]. In comparison, van Drongelen et al. [3] measured muscle strength subjectively as scored by the research assistant. Aljure et al. [16] focusedecifically on the incidence of CTS and assessed this by utilizing electrophysiological studies of the median and ulnar nerves following a standardized protocol [34].

4.13 Prevalence

All studies reported various areas and levels of upper limb pain or injury. Detailed prevalence rates by setting have been outlined in Table 3. The most common area of pain reported in the upper limb was the shoulder and the highest prevalence of shoulder pain was 71% [18], unspecified upper limb pain reported was 81% [21], however it is not aligned to any particular structure of the upper limb.

Setting*	Measure	Shoulder	Elbow	Wrist	Hand
Inpatient setting N = 2	Median	33%	N/A	N/A	N/A
	Mean	33%	N/A	N/A	N/A
	Highest	56.5%	N/A	N/A	N/A
	Lowest	39%	N/A	N/A	N/A
Outpatient setting N = 4	Median	66%	25.5%	33.5%	28%
	Mean	61%	25.2%	33.5%	28%
	Highest	71%	35%	53%	43%
	Lowest	41%	15.5%	14%	13%
Community setting N = 4	Median	39%	20%	40%	45%
	Mean	54.3%	20%	33.8%	45%
	Highest	73%	31%	55%	45%
	Lowest	35.6%	9%	6.6%	45%
Community and outpatient N = 2	Median	31%	N/A	N/A	N/A
	Mean	31%	N/A	N/A	N/A
	Highest	32%	N/A	N/A	N/A
	Lowest	30%	N/A	N/A	N/A
Range and mean of combined prevalence estimates:		Shoulder	Elbow	Wrist	Hand
		Range = 35.6% - 73%; Mean = 44.8%	Range = 9% -35%; Mean = 22.6%	Range = 6.6% - 55%; Mean = 33.6%	Range = 13% - 45%; Mean = 36.5%

Table 3.

Prevalence of upper limb pain by setting.

^*

[16] did not detail the setting of their study and no specific percentages of pain per area were reported in [3], both have therefore been excluded from this table.

Prospective cohort studies were undertaken in 5 studies [3, 13, 14, 18] and they recorded the level of upper limb pain. Ballinger et al. [13], reported an increase of shoulder pain over the 3-year study, and this was more prevalent in men who were older, reported poorer health and had acromioclavicular (AC) joint narrowing as determined by X-ray on first admission to rehabilitation. In contrast to this, van Drongelen et al. [3] reported a decrease in shoulder pain (30%) at the second test point. Muscle strength was significantly inversely related to shoulder pain at the beginning of rehabilitation and body mass index (BMI) was a strong predictor for pain, one year after in-patient rehabilitation. Both [13, 14] reported 32% of participants had limited shoulder ROM and 39% reported pain at the shoulder on discharge from rehabilitation.

Aljure et al. [16] and Escobedo et al. [20] both investigated the prevalence of a specific injury, CTS and RCT respectively. Aljure et al. [16] reported 63% of participants had electrical nerve abnormalities confirming the presence of CTS, while 44.7% also had ulnar nerve neuropathy. Escobedo et al. [20] reported 70% of participants were symptomatic of RCT, with MRI imaging showing 62% full RCTs and 12% partial RCTs. Samuellson et al. [24] was the only study to associate pain with a diagnosis of a condition. Thirty seven percent of participants reported shoulder pain, with findings of muscular atrophy, pain, impingement, and tendinopathy described. The estimated mean prevalence of upper limb pain by outcome measure has been detailed below in Table 4.

	Questionnaire element N = 6	Physical Exam N = 3	Radiographic element N = 3	Electrophysiological element N = 1
Mean	52.6%	50.3%	50%	63%
Highest	71%%	73%	70%	63%
Lowest	35.6%	39%	30%	63%

Table 4.

Estimated mean prevalence of upper limb pain by outcome measure.

4.14 Relationship of pain with participant characteristics

The relationship between pain and wheelchair user’s characteristics was investigated in thirteen studies. Significant results were reported in nine of these studies. Time since injury was a significant factor in predisposing participants to the development of upper limb pain. More specifically, [15, 21, 22] reported the development of unspecified upper limb pain was significantly associated with length of time since injury and Aljure et al. [16] reported significant incidence of CTS increased with length of time since injury. Level of SCI was significantly related to upper limb pain in two studies [2, 22]. Pentland & Twomey [22] reported pain is significantly associated with participants with paraplegia compared to the able-bodied population, while van Drongelen et al. [3] reported participants with tetraplegia are significantly predisposed to developing upper limb pain compared to participants with paraplegia.

Two studies reported significant relationships between upper limb pain and age [18, 20] although this contradicts findings from three studies who reported no significant correlation between pain and age [22, 24, 26]. Additionally, radiographic results from Boninger et al. [17] found a significant relationship between imaging abnormalities and Body Mass Index (BMI), but not pain.

4.15 Relationship of pain with functional activities

Of the studies reviewed, eight assessed the impact of upper limb pain on functional activities. Dalyan et al. [18], reported the highest level of pain was associated with pressure relieving, transfers, and wheelchair mobility. Gironda et al. [21] similarly to Dalyan et al. [18], reported wheelchair mobility and transportation as the activities resulting in the greatest amount of pain in the upper limb. Further to this, El Essi et al. [19] examined wheelchair mobility to include pushing a wheelchair, propulsion up ramps and outdoor inclines as the primary contributors to upper limb pain. Seventy-four percent reported no limitation during recreational or athletic activities, while the remainder agreed that pain had limited function to varying degrees. Few participants reported seeking treatment for this issue, only 23–35% made changes to their routines and 6–16% had sought assistance from a carer or friend with ADLs due to upper limb pain.

Samuellson et al. [24], used the Canadian Occupational Performance Measure (COPM) to assess the impact on ADLs. From this, issues in 52 areas of occupational performance were associated with upper limb pain, with 54% of these related to self-care. Furthermore, van Drongelen et al. [3] found upper limb pain to be significantly inversely related to functional outcome. Eriks-Hoogland et al. [14] reported limitations of shoulder ROM were significantly associated with the ability to transfer, FIM motor scores and participants returning to work. Pentland & Twomey [22] devised their own questionnaire based on the Barthel Index. Although functional limitations were not formally assessed, participants with pain reported tasks most impeded by pain included work/school, sleep, wheelchair transfers, outdoor wheeling, and driving.

One study included a female only sample [23], participants reported outdoor wheeling as the most difficult task to complete while experiencing pain. Additionally, Ballinger et al. [13] reported men with shoulder pain scored lower CHART and FIM scores, however, this was not statistically significant.

4.16 Study quality appraisal

Appendix 2 provides details on the quality of the studies. There were four cohort studies and eleven cross-sectional studies. The cohort studies scored moderately well on the checklist with all scoring positively on over half of the criteria [3, 14, 15, 18]. The remaining cross-sectional studies scored lower overall due to several biases relating to study design and analysis of data. In relation to the studies composed of a radiographic element, only one study blinded the reporting radiographer to participants [17]. Escobedo et al. [20] stated three observers interpreted the MRI results however it is unclear if they were blinded or what level of expertise they held. Four studies did not include any standardized outcome measures therefore questioning the validity and reliability of their results [18, 22, 23, 25]. Self-reporting questionnaires are also a limitation as they are likely to present an over endorsement bias, where participants answer questions relating to their health in an enthusiastic manner, often over reporting the extent of their pain or injury [35].

Physical assessments were conducted in twelve studies with five of these studies following standard protocols for the reporting of muscle strength and ROM [3, 15, 16, 24, 26]. Although, van Drongelen et al. [3] used a standardized protocol to conduct manual muscle testing, muscle force was subjectively measured by the research assistant therefore impacting the quality and objectivity of results reported. Pentland & Twomey [22, 23], were the only two studies to use mechanical devices to measure muscle strength via use of a dynamometer. Dynamometers are well documented as accurate devices in reporting grip strength and therefore add to the methodological quality of these studies [36].

Sample size varied greatly across all studies. A larger sample size increases the validity of results as it reduces the chance of error that results occurred because of another reason and not the hypothesis in question. Four sample sizes included over one hundred participants however it was unclear if power calculations were conducted to ensure generalizability of results. The smallest sample sizes were observed in [23] 11 participants and [17] 28 participants. A smaller sample size increases the risk of error in applying results to the wider SCI population and therefore these results should be interpreted with caution.

Recruitment bias refers to the methods utilized by studies for inclusion of participants. Several studies recruited participants from specific hospitals catering for different diseases or conditions. Five studies recruited participants from Veteran Affairs Hospitals [13, 17, 20, 21, 26] who provide care specifically to Veterans and their families. Recruitment bias may exist where participants may not be an accurate representation of the wider SCI population, or it may result in an uneven representation of the wider population as the hospital caters to a specific population of SCI patients.

4.17 Causation of secondary musculoskeletal (MSK) injuries

The etiology of upper limb pain was primarily attributed to the overuse of the upper limb during wheelchair propulsion and transfers in twelve studies. Functional activities which exacerbated pain the most included outdoor wheeling, ramps/inclines, wheelchair transfers and domestic ADLs (DADLs). Gironda et al. [21] concluded that although the overuse of the upper limb contributed to injury or pain, it was not sufficient in explaining the development of pain itself. They stated the development, persistence and exacerbation of pain is further aggravated by functional activities; however, injuries would be best understood in the context of a theoretical model to understand the person as a whole.

Similarly, Subbarao et al. [26] reported that not all pain can be attributed to the overuse of the upper limb alone. They reported that acute trauma to a joint or structure in the upper limb could cause early pain, while cumulative trauma may result in late onset of injuries. Incorrect loading of joints or abnormal movement patterns were viewed as the primary causation factors of upper limb pain in two studies [15, 24].

Samuelsson et al. [24] discussed the anatomical positioning of wheelchair users during wheelchair propulsion. He concluded the kyphotic position wheelchair users adopt while propelling places further strain on the shoulder joint, depressing the acromial process, and changing the facing of the glenoid fossa, thus resulting in pain and injury. Similarly, Silfverskiold & Waters [15] attributed the causation of injury to abnormal glenohumeral motion during active or passive ROM of the shoulder joint. Boninger et al. [17] was the only study to attribute the causation of pain to increased BMI in SCI participants. They reported an increased BMI resulted in increased weight for participants during wheelchair propulsion and transfers, thus placing further strain on the upper limb joints and structures.

Distinctly, only two studies attempted to distinguish the type of pain experienced by participants. Neuropathic pain is a common occurrence in the SCI population where pain occurs below or surrounding the level of injury. Both Eriks-Hoogland et al. [14] and van Drongelen et al. [3] attempted to distinguish between neuropathic pain and upper limb pain. Both used self-reporting questionnaires advising participants to report only pain they experienced because of trauma or injury, not directly related to their injury. It is not always possible to distinguish between both types of pain and the use of self-reported questionnaires placed the onus on participants to decipher this individually. It is therefore difficult to confirm if pain that was neuropathic in origin was included in their analysis.

4.18 Treatments sought

Only four studies reported on treatments availed of by participants experiencing upper limb pain [18, 21, 22, 25]. Dalyan et al. [18] provided the most in-depth detail relating to treatments, stating 63% sought medical intervention on experiencing pain. Of this, 90% received either physiotherapy, pharmacological treatment or massage, and home modifications or joint protection education was sought by 27% of participants. Joint protection education was reported to be most beneficial by 63.3% of participants, however it is unclear when, or who delivered this. Twenty-six percent of participants also found home modifications useful.

Both Gironda et al. [21] and Sie et al. [25] detailed how 43% and 30% of participants respectively used opiate medications daily, which provided only moderate relief. Pentland & Twomey [22] discussed treatment options availed of by participants and found that many participants were fearful of seeking treatments such as steroid injections, surgery, or hospital admission due to the invasive nature of such. The final treatment option which was discussed was that of resting the upper limb, however participants felt this was unachievable.

5. Discussion

The results from this systematic literature review highlight varying prevalence rates of upper limb pain across 15 studies. The shoulder was the primary pain site investigated by studies, with three studies investigating prevalence of pain of the upper limb in its entirety.

Prevalence rates ranged from 11–81% and differed by reporting measures, outcome measures utilized, recruitment methods, level of injury of participants, time since injury and age. Little is currently known regarding prevalence rates of upper limb pain in SCI, however it is anticipated this review will highlight the variety of research undertaken and gaps in knowledge relating to upper limb pain in the SCI population.

There was considerable variation in the method of data collection across all studies. The heterogeneity of studies implies difficulty in drawing overall conclusions from the studies included [37]. The reported pain values vary from 11% -81%; no clustering of prevalence rates was noted suggesting the samples are heterogeneous. The varying levels of SCI were not consistently recorded. Some studies used the ASIA scale, some studies stated either participants with tetraplegia or paraplegia, and some studies stratified participants based on the medical level of injury reported. The lack of standard criteria defining level of injury in each study offers minimal help in explaining between-sample differences thus making it difficult to report results applicable to the wider SCI population.

The use of self-reported questionnaires was the most prevalent methodology utilized on the basis that they are cost effective and easy to administer. Self-reported questionnaires have been used widely across healthcare research to obtain prevalence rates, health status and health services accessed [38]. Self-reported questionnaires are useful when the data required is not normally collected via audits or medical practice or when database analysis is deemed too expensive or time consuming to conduct [39]. Despite the widespread use of these, there is little consensus regarding the accuracy of information reported and the validity of findings [40]. Potential bias lies in the over or under-reporting by participants such as recall timeframe where participants may suffer memory decay. Literature shows an increased number of hospital or healthcare visits results in an under-reporting of the number of visits; the more often they occur, the less memorable they are to participants [41, 42, 43]. Over endorsement bias may also exist where participants may under or over-report pain to please their healthcare professional or as an incentive to be included in a research study. Although this questions the validity of results, self-reported questionnaires are often the only option to obtain data when it is not recorded elsewhere.

A systematic review investigated prevalence of pain in cancer patients [44]. They found the use of self-reported measures were more reliable than medically documented symptoms, as pain was only recorded by 10% of oncologists, resulting in the underestimation of the prevalence of pain. This is in part due to the complex nature of cancer where pain may not have been a priority for the physician to assess. It is reasonable to draw comparisons between the recording of pain in cancer populations and SCI populations as both conditions are complex in nature and potentially have more critical issues associated with their condition to report. Individuals who are diagnosed with a condition or illness are also less likely to report abnormal sensations or health related issues as they attribute these to the disease itself [45].

Muhajarine et al. [46] conducted a study on individuals with hypertension and compared the efficacy of self-reporting questionnaires to that of an able-bodied population. They reported that participants with hypertension were less likely to report abnormal issues via use of a self-reported questionnaire in comparison to attending a physical assessment by a healthcare professional. Similarly, to patients with an SCI, it could be argued that they felt this complaint was not significant enough to formally report in a questionnaire, however a face-to-face consultation may identify pain via a physical assessment or may allow healthcare professionals to probe further during consultations.

Within this current review, three studies utilized radiographic imaging to explore the pathology of pain and three studies also invited participants to attend for a physical assessment of their pain. The variance in methodology may have contributed to the variance in prevalence rates reported. A physical exam by a trained healthcare professional may provide objective reporting of injuries however a lack of standardized outcome measures utilized by studies resulted in data lacking validity and reliability.

Physical assessments of pain may also be deemed as invasive for participants who experience pain, and an additional burden lies on the participant in attending appointments and undergoing tests for the purpose of a research study. For the research team, both the use of physical assessments and radiographic imaging are time consuming and require expert knowledge and a number of assessors in order to ensure reliability and validity of results. Taking all of the above literature into account, the use of a self-reported questionnaire in the SCI population is feasible and cost effective, however may not be sufficient in accurately reporting the prevalence of pain or treatments availed of. Therefore, it could be argued that participant reported prevalence rates could be confirmed by accessing patient medical notes to determine specifically what pain they reported, how often it was reported, and treatments prescribed for the management of their pain.

The reporting of pain may also lead to questions around the validity of results in this review. Research evidence shows that of those with SCI who have experienced chronic pain, 40% of patient’s pain is neuropathic in origin [47]. Neuropathic pain (NP) can occur above, at, or below the level of SCI and is commonly described as sensations of “burning”, “stabbing”, or “electric shock like” [2, 48]. Given the expressed unsettling and untreatable nature of the pain by the patients themselves, it is not surprising that NP is one of the most frequently reported and most difficult to treat secondary health conditions associated with SCI [49]. The chronicity and prevalence of pain is strongly associated with an increase in hospital visits and utilization of medical services [50, 51] NP is also quite difficult to distinguish from musculoskeletal pain. NP can occur at or below the level of injury, however in incomplete SCI, MSK pain can also occur at these sites thus making it difficult to determine the origin of pain.

Within this review, only two studies defined the origin of the type of pain experienced. Although some studies linked pain experience to functional activities, it is difficult to decipher whether the pain experienced is related to the level of injury or whether the pain is from functional activity alone [52]. The use of self-reported outcome measures further confounds this, putting the onus on participants themselves to make this distinction, which may prove difficult.

The causation of pain was attributed to the overuse of the upper limb in twelve studies. Wheelchair users rely on the upper limb for mobilizing on a daily basis so it is unsurprising that this plays a role in the development of pain. Two studies referred to the development of pain stemming from anatomical positions adopted during specific wheelchair related activities. With such a small number of studies reporting this, it is difficult to determine if this is the sole source of pain or if there are other variables involved. Further research relating to the biomechanical movement patterns of wheelchair use may help explore the etiology of injuries. Furthermore, wheelchair skills training could play a role in educating patients on joint protection during activities [26].

Pain was most exacerbated by outdoor wheeling, propelling up ramps or inclines and wheelchair transfers. Education around energy efficient propulsion techniques or use of assistive technology to aid transfers may prove beneficial, however there is little literature to confirm this. Only four studies discussed the type of treatments the participants availed of. Only one study [22] further investigated the use of treatments and found participants were fearful of seeking invasive treatments for relief, and rest was deemed unachievable. The question remains, what treatments are available, what are the advantages/disadvantages of each and how effective are they at relieving pain? Further research is also required to understand the implications of pain for participants. How does pain affect their day to day lives with work/school activities, sleep, personal care tasks, domestic ADLs, childcare, or other psychosocial elements of their lives.

To the author’s knowledge, only one study from the United Kingdom (UK) has addressed the prevalence of upper limb pain in the SCI population. Nichols et al. [53] was one of the earliest studies to document the phenomenon of overuse injuries in the SCI population, however, was excluded from this review on the basis that powered wheelchair users were included in the sample. Statistics relating to wheelchair use in Northern Ireland are limited, with the most recent figures estimating approximately 30,000 of the 1.8 million population of Northern Ireland classified as wheelchair users [54]. This equates to 1.3% of the Northern Ireland population which is below the UK National average of 2%. It is not clear how accurate the regional figures are, and they may not reflect the true situation. Northern Ireland has a strong history of conflict, most noticeably “The Troubles” which lasted from 1960 to 1998, resulting in over 47,000 individuals injured and 500 severely injured [55].

Indeed, a similar country with a history of conflict (but on a greater scale) took place in the Gaza Strip, Israel [19]. They hypothesized that the number of persons with an SCI in the Gaza Strip increased due to the conflict during the Al Asqa Intifada (2000–2005). Excessive force and the use of explosive devices was prevalent in war torn areas resulting in widespread casualties. Similar to El-Essi et al. [19], it is reasonable to argue that the number of wheelchair users or those with an SCI is potentially under-reported in Northern Ireland. From 1960 to 1998 there were 36,923 shootings, 16,209 bombings and approximately 47,541 people were injured in Northern Ireland (Conflict Archive on the Internet last modified 1/02/18). Those who may have been injured during the troubles 10–50 years ago are now long-term wheelchair users. With length of time since injury significantly associated with the development of upper limb pain, and a potential greater sample of wheelchair users in Northern Ireland as a result of The Troubles, it is reasonable to hypothesize that Northern Ireland will have a higher SCI population and specifically a higher percentage of upper limb pain as documented in long-term wheelchair users. There is currently no literature documenting the prevalence of upper limb pain in the SCI population of Northern Ireland, a significant gap in knowledge considering the history of the country.

5.1 Review limitations

This review was limited in that only studies specifically referring to upper limb pain were included. Studies reporting on generalized pain in the SCI population were excluded as they were not directly relevant to the research question. Other limitations of the study were due to the exclusion of studies not written in the English language. Studies specifically focused on wheelchair athletes were also excluded as this population experience a higher level of physical activity and the potential for sporting injuries may skew results rather than reporting of injuries sustained by manual wheelchair use alone.

6. Conclusion

The increasing number of people with an SCI living longer and healthier lives comes with a consequence of secondary musculoskeletal impairments. The most common site of pain investigated was the shoulder. Varying reporting measures of age, time since injury, level of injury and standardized outcome measures hampered the comparison of the overall prevalence rates of upper limb pain. Little is currently known of the etiology of upper limb pain, treatments available for upper limb pain or how pain affects sufferers on a daily basis. A uniform measurement of upper limb pain specific to the SCI population would be useful in comparing prevalence rates, however none currently exist. A basic pain data set (International Spinal Cord Injury Basic Pain Data Set, ISCIPDS) has been developed within the framework of the International Spinal Cord Injury data sets with the purpose of facilitating consistent collection and reporting of pain in the SCI population [56, 57] however, it is not specific to the reporting of upper limb pain. Future research should focus on what treatments are available and most effective at treating upper limb pain in SCI, specifically in Northern Ireland where an underestimated population of long-term wheelchair users may exist.

Conflict of interest

The authors declare no conflict of interest.

Reference	Design	N =	Aims	Follow up	Method	Outcome measures	ADLs
[16]	Cross sectional	47 (91 hands)	To assess the prevalence of carpel tunnel syndrome in patients with paraplegia	N/A	Electro physiological studies of the median and ulnar nerves; Physical exam	Standardized protocol for conducting tests according to Johnson 1980	No
[15]	Cohort	89	To determine if shoulder pain and ROM problems can be predicted by demographic, injury related, body weight and radiographic data over 3 years	3 years	Radiographic assessment of shoulders in anteroposterior position; Questionnaires Physical exam	FIM CHART	Yes
[17]	Cross sectional	28	To use magnetic resonance imaging (MRI), plain radiographs, questionnaire and physical exam to gain insight into the prevalence of shoulder disorders	N/A	MRI; X-ray Questionnaire Physical exam	MRI clinical protocol for identification of rotator cuff tears (RCT); X-rayed in AP, scapular AP and supraspinatus position	No
[18]	Cross sectional	130	To determine the frequency and severity of UE pain and its association with functional activities	N/A	Postal questionnaire	Non-validated questionnaire	Yes
[19]	Cross sectional	80	Examine the prevalence of shoulder pain and its effects on ADLs	N/A	Interview Questionnaires	WUSPI; Shoulder Rating Questionnaire (SRQ)	Yes
[14]	Cohort	138	Examine whether MSK shoulder pain at first discharge are associated with ADL restriction at 5 years	5 years	Questionnaire Physical exam; 3 wheelchair related tests	Physical Activity Scale for Individuals with Physical Disabilities (PASIPD); 2 subscales from the sickness Impact Profile 68 (SIP68) (mobility range and; social behavior scales)	Yes

[20]	Cross sectional	37	To use MR imaging to evaluate the prevalence and extent of rotator cuff tears in patients with paraplegia	N/A	MRI (3 observers to interpret results); Scanned patients in supine position with arms adducted and the humerus head in; neutral	MRI machine (Gyroscan ACS-2)	No
[21]	Cross sectional	669	Examine the prevalence and intensity of pain and associated patient characteristics in paraplegia	N/A	Questionnaire of medical history Questionnaire	WUSPI	Yes
[53]	Cross sectional	517		N/A	Postal survey		No
[22]	Cross sectional	52	Describe the effects of longterm paraplegia and wheelchair use on upper limb function	N/A	Physical exam – physical performance and parameters of upper limb function Interview Questionnaire	Upper extremity isokinetic and grip strength, pain and active ROM using KinCom II isokinetic dynamometer	Yes
[23]	Cross sectional	11	To compare upper limb function and pain in wheelchair using women with paraplegia to a matched able bodied sample	N/A	Physical exam – physical performance and parameters of upper limb function Interview Questionnaire	Upper extremity isokinetic and grip strength, pain and active ROM using KinCom II isokinetic dynamometer Smedley’s hand held dynamometer	Yes
[24]	Cross sectional	56	To describe the consequences of shoulder pain on activity and	N/A	Questionnaire WUSPI Interview	WUSPI; Constant Murley Scale Klein and Bell ADL index	Yes

			participation in SCI wheelchair users with paraplegia		Constant Murley Scale Klein and Bell ADL index COPM; Physical exam	COPM
[25]	Cross sectional	239	To determine the prevalence of upper extremity pain in outpatients with SCI	N/A	Questionnaire Interview Physical exam; (Pts offered physical exam; following pain)	2 point discrimination and Semmes- Weinstein monofilament testing	No
[15]	Cohort	60	To determine the incidence of non-traumatic shoulder pain and associated functional disability during the first 18 months after SCI	18 months	Physical exam following standard protocol at 6 months and then between 6 and 18 months following this Functional disability; questionnaire	Own questionnaire Physical exam	Yes
[26]	Cross sectional	451	To identify the prevalence of chronic wrist and shoulder pain, to determine which activities caused or exacerbated pain and assess functional and emotional responses and how pain might be reduced.	N/A	Review of medical records Postal survey; Physical exam (n = 30) Interviewed prior to physical exam; Included completing functional tasks transferring, propelling and dressing; upper bodies	Own questionnaire previously pilot tested; If pain reported in questionnaire participants were interviewed and then physical exam using standardized evaluation sheet	Yes
[3]	Cohort	169	To study MSK UE pain during and after rehabilitation in wheelchair using participants with SCI and its relationship with lesion characteristics, muscle strength and functional; outcome	1 year	4 test occasions Physical exam; MSK pain questionnaire Manual muscle testing (MMT)	Lesion and personal characteristics assessed by physician; Used standardized questionnaire MMT conducted in standardized positions; Muscle force measured subjectively by research assist.; FIM	Yes

ROM = Range of Movement; FIM = Functional Index Measure; CHART = Craig Handicap Assessment and Reporting Technique; MRI = Magnetic Resonance Imaging; RCT = Rotator Cuff Tear; AP = Anteroposterior; UE = Upper Extremity; ADLs = Activities of Daily Living; SRQ = Shoulder Rating Questionnaire; PASIPD = Physical Activity Scale for Individuals with Physical Disabilities; SIP68 = Sickness Impact Scale; WUSPI = Wheelchair Users Shoulder Pain Index; SCI = Spinal Cord Injury; COPM = Canadian Occupational Performance Measure; MSK = Musculoskeletal; MMT = Manual Muscle Testing.