Introduction

Although loss of function is considered the most significant consequence of spinal cord injury (SCI), pain is a debilitating accompaniment that imposes significant burden on individuals who have already suffered substantial emotional and physical trauma.1, 2 Pain in general is defined by the International Association for the Study of Pain (IASP) as ‘an unpleasant sensory and emotional experience associated with actual or potential tissue damage’.3 SCI-related pain is heterogeneous; several subtypes are presumed to exist, each with different pathophysiology and likely different treatment approaches.4, 5 The most common of these in SCI are neuropathic pain elements, defined as ‘pain arising as a direct consequence of a lesion or disease of the somatosensory system’.6

Data concerning the prevalence, causes, characteristics and treatment of chronic pain in the SCI literature are accumulating but are not yet definitive.7, 8 Prevalence rates for SCI-related pain range between 48 and 94% of the SCI population, depending on population characteristics (for example, acute, chronic) and measurement factors (for example, pain intensity, interference).9, 10, 11 However, regardless of prevalence, pain in SCI is remarkable for its chronicity, interference with functioning and resistance to medical treatment.12, 13

A lack of standardization and agreement regarding both classification and description of pain elements hampers understanding of pain in SCI, especially for neuropathic pain elements.14, 15 Historically, clinical assessment of pain has generally followed a commonly utilized, but loosely applied format focused on subjective phenomena of pain.16 However, pain is ‘essentially a subjective experience described by person-specific symptoms and expressed with a certain intensity’.17 Measurement must therefore be reliable, valid and responsive to pain phenomena;18 a standardized and reliable classification is essential. Determination of pain type, prevalence and associated complications should lead to specific, circumscribed treatment approaches and permit meaningful comparisons across studies.4

Although a considerable volume of literature exists on pain in the SCI population, there is a notable lack of adequate documentation on the psychometric properties of instruments used specifically in SCI pain research.19 Selection of appropriate pain-related instruments for clinical or research purposes is therefore uncertain.4 To address this deficiency, this critical review assesses the psychometric properties (reliability, validity, purpose, utility and responsiveness) of measures commonly utilized in SCI pain research.

Three aspects of pain measurement were addressed: classification of SCI-related pain (pain description according to type, region and presumed mechanism), pain perception (intensity, sensory description) and pain interference (impact on daily life/activity).

Methods

PubMed, MEDLINE, CINAHL, EMBASE, HaPI, PsycINFO and SPORTDiscus electronic database searches were undertaken to identify peer-reviewed articles published between 1986 and 2006 that reported on pain measurement in SCI. Additional searches were conducted through archival of references in papers obtained by electronic search. The keyword ‘spinal cord injury’ was used in all database searches and then combined with other terms (that is, pain, pain measures, pain measurement, reliability, validity, psychometrics, reproducibility of results and data collection) depending on database parameters.

Approximately 1000 peer-reviewed publications that addressed SCI pain were identified. Article titles and abstracts were reviewed for initial retention of relevant articles on SCI pain. Sixty-seven articles specifically addressed measurement of SCI pain, describing various algorithms or measures for evaluation of a wide array of pain-related domains and the impact of pain on functioning. To be included, measures had to have psychometric properties assessed and reported in at least one peer-reviewed study using a SCI population. Selected papers were carefully reviewed and relevant psychometric information was extracted. (For the purpose of this review, RCT studies were not included since the primary focus of this review was methodological analyses of the tools. While reliability (and responsiveness—effect size can be calculated) and statistics can be calculated in these studies, the formulae used are often different.)

Evaluation procedure

A review team consisting of clinicians and scientists with established expertise in research areas relevant to SCI issues was established. The team reviewed each of the articles as part of a larger review of health care and rehabilitation strategies for persons with SCI.20 Data about psychometric properties of the identified measures (that is, reliability coefficients, type and magnitude of validity correlations, responsiveness) and ‘pragmatic’ factors (that is, interpretability, acceptability and feasibility) were extracted from manuscripts in accordance with similar projects designed to review outcome measures.21, 22, 23

Selection criteria

Standards for quality of reliability, validity and responsiveness for each measure were adapted from a previously published overview of outcome measurement criteria.24 Reporting format for rigor and quality replicated the approach adopted by McDowell and Newell25 adapted for the SCI population. Definitions, cutoff ratings and criteria details for outcome ratings are briefly summarized in Table 1 and further explicated by the SCIRE authors.20

Table 1 Criteria for rating outcome measures

Results

Seventeen studies identified by this process reported psychometric properties of measures in studies of SCI pain. Five SCI pain classification systems were identified. Eleven studies were identified that evaluated seven self-report measures of pain perception. Four self-report measures of pain perception and two self-report measures of pain impact on functioning were subjected to review (Table 2).

Table 2 Summary of psychometric properties

Pain classification

Stated goals were similar across the five SCI pain classifications, namely to standardize terminology, provide direction for treatment, inform outcome evaluation and promote knowledge translation of basic science advances to clinical practice. Early constructions (Tunks,26 Donovan27 and Siddall7 classifications) were instrumental in providing initial categorizations for SCI pain. These early constructs are superceded by three recent SCI pain classification systems. Siddall's initial construction, modified to a three-tier taxonomy, is most well known as the IASP Task Force on Pain following SCI.28 This schema classifies SCI pain according to type (nociceptive and neuropathic), subtype (musculoskeletal, visceral at, below and above lesion level) and finally presumed mechanism (specific structure/pathology) (Table 3).

Table 3 Pain measures meeting study criteria

The most recent SCI pain categorization construction (Cardenas et al.29) is comprised of two major categories (neurologic and musculoskeletal pain) further divided into four neurologic pain subcategories (SCI, transition zone, radicular and visceral). SCI level of lesion, level of pain (at, below and above lesion level), pain laterality, responsiveness to pain stimuli/activity and SCI completeness are also used as criteria for categorization purposes in this multi-axial assessment protocol. Other classification systems have also been published, such as the Bryce/Ragnarsson-SCI-Pain Taxonomy (BR-SCI-PT),30 a three-tier system aligned with Siddall's original classification by level of injury, pain type (nociceptive or neuropathic elements) and subtype (regional localization) and Siddall's original model.7

Five pain classification schemes met our inclusion criteria: Tunks Classification Scheme,26 Donovan SCI Pain Classification,27 Siddall Classification,7 IASP Taxonomy of SCI Pain,28 BR-SCI-PT30 and Cardenas’ Pain Classification29 (Table 4). Each addressed presumably differing underlying pathologies for different subtypes of SCI pain, pain location, clinical presentation and temporal patterns. Adequate but not excellent evidence supporting the quality of reliability as indicated by the magnitude of the coefficient inter-rater reliability was reported for each of these schemas, however, no studies investigating empirical elements of validity of these classifications were identified. The models are difficult to compare given varying formats, number of items and rating scale parameters.

Table 4 Pain classification schemas

Measurement of self-reported pain

A total of 11 papers investigating measurement issues on 6 different self-report pain scales were identified. These tools were primarily developed for use in other populations. The majority of psychometric studies on measurement of pain sensation in the SCI population reported only reliability data, again predominantly test–retest reliability (Table 4). The McGill Pain Inventory for SCI reported ICC values ranging from 0.69 to 0.78.31 The Wheelchair Users Shoulder Pain Index (WUSPI) had the highest reliability measures (ICC=0.99).32, 33 The Brief Pain Inventory (BPI)34 and the WUSPI had the highest values for internal consistency with Chronbach's α all greater than 0.91.35 Of the few validity studies identified, most assessed either convergent or content validity; measures were statistically significant with BPI demonstrating the highest correlations.35

Pain intensity

Most studies on SCI pain utilize some form of one-dimensional rating scale for evaluation of self-reported pain intensity (for example, magnitude or severity). Numerical rating scales (NRSs) are the most commonly applied metric for pain report in both the general and SCI pain literatures for subjective estimates of pain experience. NRSs are well-established for validity, application facility and responsiveness; these are the most commonly applied metric for pain report in SCI research studies.36, 37, 38 The most utilized constructions are 11-point Likert Scales with anchors at ‘0’ (no pain) and ‘10’. Upper-end anchor descriptors vary across studies (for example, pain as bad as it could be, the worst pain imaginable). Most studies assessed in this review employed NRS items of varying anchor descriptions.

Pain sensory symptoms

Measures developed for general chronic pain populations that include various sensory (verbal description) and/or affective (reflective of pain-related distress) descriptors of pain quality are commonly used in SCI pain research (see Appendix 1 for descriptions of the measures included for review). The most frequently reported measure of sensory quality in the SCI literature appears to be the McGill Pain Questionnaire or its' variations.3, 11 The Medical Outcomes Survey-Short form (MOS; SF-12),39 Graded Chronic Pain Disability Scale (GPS)40 and Multidimensional Pain Inventory (MPI)41 have also been employed in the SCI pain literature (MPI-SCI)42 and report solid reliability psychometric data.

The majority of psychometric studies on measurement of pain sensation in SCI populations reported only reliability data, predominantly test–retest reliability (Table 4). The MPI-SCI reported ICC values ranging from 0.69 to 0.78.42 All of the self-report pain measures have been assessed for construct or convergent validity with other measures with adequate level for correlations. Grade Chronic Pain Scale and the Brief Pain Inventory (BPI) correlated well with the SF-36 (r=0.5 and r=0.6, respectively) while the McGill Pain Questionnaire correlated well with Sternbach's Pain Intensity Scale11 (r=0.5). The extent to which short forms (for example, SF-12 and MPQ-SF) of standard measures are adequate for providing a comprehensive picture of SCI pain perception remains unclear.24 These, and similar scales, are also restricted with regard to inclusion of neuropathic pain elements common to SCI. Three recently developed measures assessing neuropathic pain in general medical conditions are reported in the general pain literature but the range of neuropathic symptom descriptors appear limited and none are specific to SCI pain experience.17

Pain interference

Pain interference measures purport to assess the nature and degree to which pain negatively impacts various aspects of functioning. Such measures using NRSs or Likert-type items are commonly used in descriptive and treatment outcome evaluation studies in general, and SCI, pain literature.43, 44, 45 Most studies addressing SCI pain include some form of pain interference items using various modifications of the Brief Pain Inventory (BPI),34 either independently or in conjunction with other measures. However, variability is the norm for item inclusion and format although most utilize at least one item on pain interference that addresses the ability to participate in customary activities. Anchor descriptors vary (for example, unable to carry on any activities, completely interferes) and are commonly modified to better reflect SCI experience (for example, ‘mobility’ vs ‘walking’). As others have pointed out, adaptations to measures such as these for disability populations should be examined in concert with review of factor structures to ensure face validity and scale integrity.24

Only one published measure constructed solely for SCI pain interference was identified. The Wheelchair Users Shoulder Pain Index (WUSPI)32 appears to be a psychometrically sound instrument that assesses the impact of shoulder pain in persons with SCI who are wheelchair reliant for mobility.33 The WUSPI reported high reliability values (ICC=0.99) and showed adequate validity with functional upper extremity range of motion measures (r=0.47). Both the BPI and WUSPI reported strong internal consistency (Cronbach α's all greater than 0.91).

Discussion

This study provides information about published psychometric data on pain measures used to evaluate pain issues in persons with SCI. Despite improvement in the quality of SCI literature over the past two decades, studies on SCI pain measurement are limited compared to pain studies in other populations.26, 46 Very few published studies that reported psychometric properties of pain measurement specific to the SCI population were identified by our search criteria. Variability in the application of clinical and research tools for SCI pain assessment and measurement was evident. This lack of uniformity precludes comparison across SCI pain studies.4

Which classification system and pain measures should be used? A number of models have been proposed for the classification of SCI pain. The most recent of these (IASP Taxonomy of SCI Pain,28 Bryce-Ragnarsson SCI Pain Taxonomy,30 Cardenas Model29) focus on pain description according to region (relative to level of injury) and underlying pathology to differentiate between nociceptive, neurologic and visceral pain. Goals are similar across schemas, namely the improvement of assessment and pain management direction of SCI-related pain. However, with the exception of universal agreement on major pain subtypes (nociceptive/musculoskeletal, neuropathic/neurologic and visceral), nomenclature and inclusion of pain types and elements are variable.

Direct comparisons of each of these models would be useful to determine comparative validity and reliability. Although these models provide some reliability data, validity remains a significant issue for pain classification, complicated in part by limitations in determining pathological mechanisms for pain origin. However, as Siddall has pointed out, for any taxonomy to be useful, it must be consistently applied.24 Universal acceptance and adoption of any one comprehensive SCI-related pain classification taxonomy remains uncertain. Although the IASP model appears to be the most utilized according to the literature, the need for specific expertise with SCI and complex pain assessment may restrict its general application.

Modifications to this model are anticipated given the recent introduction by the classification subcommittee of the IASP Neuropathic Pain Interest Group of a new definition, classification and grading system (definite, probable and possible) for neuropathic pain.6 Until an impending updated version is published, researchers and clinicians may be best served by using the current IASP schema for classifying SCI pain.

As with the general literature on pain, SCI pain studies typically report descriptive and treatment outcomes according to one of the three methodologies. The most frequent application is of single instrument use to evaluate pain-related factors such as pain intensity, pain interference/impact and pain-related disability. NRS items are almost universally employed in SCI pain studies and are supported for validity by an extensive pain literature.47

Pain intensity alone, however, assumes homogeneity of experience and has limited value in determining outcome. Clinicians and researchers are therefore encouraged to consider recommendations for structured, standard instructions and standard NRS anchor descriptions (presentation of numbers from ‘0’ ‘No Pain’ to ‘10’ ‘Pain as bad as you can imagine’) as outlined in Initiative on Methods section, Measurement and Pain Assessment in Clinical Trials (IMMPACT). IMMPACT identifies six core domains for design of clinical trials on pain: pain; physical functioning; emotional functioning; participant ratings of improvement and satisfaction with treatment; symptoms and adverse events; and participant disposition.47

With one exception, we observed that original or adapted versions of one-dimensional or multidimensional indices designed for use in populations with chronic and acute pain conditions other than SCI were most commonly used in SCI pain research. The most commonly used measures include various versions of the McGill Pain Questionnaire (MPQ),31 Graded Chronic Pain Disability Scale (GPS)40 and the Medical Outcomes Survey-Short Form (MOS; SF-12),39 all of which report satisfactory, but limited, psychometric information with SCI populations. SCI pain interference has been most commonly addressed by adaptations of the Brief Pain Inventory (BPI).34

A recently developed measure, the Pain Outcomes Questionnaire-VA (POQ-VA)48 is a multi-domain pain treatment outcome instrument that reports strong psychometric data with disability populations, including rigorous studies but unfortunately the paper did not clearly define the population outside of stating they were from the Veteran Affairs clientele who had chronic pain.17 We might wish to assume some of the subjects were from the SCI population but for the criteria outlined in this study, we excluded this paper. However, multi-method validity and reliability studies were completed on the questionnaire to evaluate the integrity of the measure. Factor analyses of the pain scales along with convergent and discriminant validity examinations in both disability and cross-validation samples were completed. This study represented the most comprehensive treatment of pain measurement in all of the measures reviewed.

Although the studies examined in this review appeared to be relatively clear with regard to stated purpose, suitability of instrument choice was at times unclear and was marked by variability in sensory description. For assessment of sensory experiences of SCI pain, no published instrument to date appears to adequately capture the complex experiences of SCI neuropathic or visceral pain. One instrument that may have possibility is the POQ-VA; a comprehensive tool that provides information on separate domains affected by pain.48 As the authors point out, this approach may be preferable for comprehensive outcome evaluation over instruments that yield isolated summary scores given variability in individual pain presentation.

Pain symptom screening tools and questionnaires, however, are naturally restricted given the lack of clarity for mechanism. However, SCI pain screening measures, at a minimum, should incorporate pain sensory items specific to and common in SCI conditions such as type (nociceptive or neuropathic); subtype (musculoskeletal or visceral) and level (at, below or above lesion). Information on thermal, dysesthetic and paroxysmal properties may also reliably be identified by SCI pain assessment tools.28 We look forward to validation studies on such instruments in the future.

There are a number of limitations to this study. While we are relatively confident that our search process captured the majority of studies meeting our criteria, database search limitations are acknowledged. Our inclusion criteria were highly restrictive and precluded studies on SCI sensory testing or randomized clinical trials, which can provide information on responsiveness.

Conclusion

Determination of which self-report pain indices are most useful for SCI pain assessment requires consideration of a variety of factors, including knowledge of the intent and properties of the measure or measures under deliberation. With few exceptions, SCI pain characteristics other than intensity, sensory descriptors of variable inclusion or impact on functioning were generally lacking in the studies examined. Knowledge of which neuropathic-type pain descriptors are most reflective of various SCI pain presentations remains unclear.

No adequate measure for the symptomatic assessment of SCI-related neuropathic pain has been developed to date. This deficiency should be addressed: development of an SCI-specific pain assessment tool that is also applicable to SCI pain research is encouraged. Temporal aspects of pain (that is, frequency, duration and time to meaningful pain relief) and pain by SCI status (tetraplegia and paraplegia) are not adequately addressed in the SCI pain literature and deserve further examination. Measures of psychological correlates, emotional functioning and global improvement and satisfaction with treatment are also lacking for standard use in the SCI pain assessment literature. The issue of meaning in pain intensity changes has also not been specifically addressed in the SCI pain literature and deserves future consideration.49

In summary, we propose a call to SCI researchers to consistently apply psychometric analyses to SCI pain data measures and look forward to increased reporting of SCI pain data and accompanying psychometric information. Greater rigor for assessing psychometric information in SCI pain studies will better inform the SCI research community of the applicability of generic measures to SCI pain investigations. Continued development and publication of psychometrically sound SCI-specific measures is also encouraged.