Abstract
Background Because of its high global burden, determining biopsychosocial influences of chronic low back pain (CLBP) is a research priority. Psychological factors such as pain catastrophizing are well established. However, cognitive factors such as working memory warrant further investigation to be clinically useful.
Objective The purpose of this study was to determine how working memory and pain catastrophizing are associated with CLBP measures of daily pain intensity and movement-evoked pain intensity.
Design This study was a cross-sectional analysis of individuals with ≥3 months of CLBP (n=60) compared with pain-free controls (n=30).
Method Participants completed measures of working memory, pain catastrophizing, and daily pain intensity. Movement-evoked pain intensity was assessed using the Back Performance Scale. Outcome measures were compared between individuals with CLBP and those who were pain-free using nonparametric testing. Associations were determined using multivariate regression analyses.
Results Participants with CLBP (mean age=47.7 years, 68% female) had lower working memory performance (P=.008) and higher pain catastrophizing (P<.001) compared with pain-free controls (mean age=47.6 years, 63% female). For individuals with CLBP, only working memory remained associated with daily pain intensity (R2=.07, standardized beta=−.308, P=.041) and movement-evoked pain intensity (R2=.14, standardized beta=−.502, P=.001) after accounting for age, sex, education, and interactions between pain catastrophizing and working memory.
Limitations The cross-sectional design prevented prospective analysis. Findings also are not indicative of overall working memory (eg, spatial) or cognitive performance.
Conclusion Working memory demonstrated the strongest association with daily pain and movement-evoked pain intensity compared with (and after accounting for) established CLBP factors. Future research will elucidate the prognostic value of working memory on prevention and recovery of CLBP.
Chronic low back pain (CLBP) is the leading cause of disability worldwide,1,2 due, in part, to its prevalence and reoccurrence.3–5 Elucidating the relationship between biopsychosocial factors and CLBP is a priority of clinical research that will assist in CLBP prevention, prognosis, and management. Pain catastrophizing is an exaggerated negative response to a painful experience6,7 and a well-established psychological distress factor associated with CLBP.8–11 Physical therapist practice standards have endorsed pain catastrophizing assessment during CLBP diagnosis and classification,12 and multiple disciplines have recommended pain catastrophizing to be a primary focus of CLBP treatment.13–15
Studies investigating cognitive factors associated with CLBP also are accumulating. Working memory is a cognitive process that links attention and long-term memory to action using phonological, visuospatial, and executive brain systems.16–19 Working memory is thought to play an important role in chronic pain because (1) attention is required to perceive pain20 and (2) anticipated pain threat and subsequent action draw from previous painful experiences.21 A recent meta-analysis supported this notion, as working memory was lower among individuals with chronic pain conditions (including CLBP) compared with healthy controls.22 Moreover, a cross-sectional study of people with chronic pain showed working memory, but not semantic or episodic memory, to be associated with pain report.23
However, unlike pain catastrophizing, potential associations between working memory and pain have not translated to clinical care. First, very few studies have assessed associations between working memory and pain, especially in a cohort with CLBP.24–26 Second, the strength of association between working memory and CLBP, compared with established factors such as pain catastrophizing, has not been determined. Studies quantifying associations between pain catastrophizing and chronic pain have accounted for other psychological distress factors, such as fear, depression, and anxiety.8,10,11 However, cognitive factors such as working memory were not included in these analyses. To our knowledge, only 2 studies have simultaneously examined working memory and pain catastrophizing in a CLBP cohort, although neither study compared the strength of association.25,27
Therefore, the purpose of this study was to assess working memory performance and associations to pain intensity among individuals with CLBP. First, lower working memory performance was confirmed among individuals with CLBP compared with pain-free individuals.22 Next, working memory and pain catastrophizing were assessed for associations with CLBP measures, both separately and through multivariate analyses, to account for age, sex, and education. Although we hypothesized that both working memory and pain catastrophizing were associated with pain,8,10,11,26 we had no a priori hypothesis regarding comparative association strength, as both factors have not been routinely assessed in parallel. The overarching study goal was to identify the unique role of working memory during CLBP, above and beyond an established measure such as pain catastrophizing.
Method
Participants
This study consisted of purposive sampling of participants with axial CLBP, stratified to a priori age group quotas of young adults (18–39 years old), middle-aged adults (40–56 years old), and older adults (57–79 years old). Age group ranges were determined based on previous research.28,29 Once a particular age group reached 20 participants, enrollment to that particular age group was closed. Our recruitment strategy ensured an equal number of participants per age group and equal representation across the life span, the lack of which is a limitation of many previous CLBP studies investigating biopsychosocial influences. Age-matched and sex-matched individuals who were pain-free (±2 years) also were included in the study. All participants were enrolled at the University of Florida Health Science Center between September 2013 and October 2014. All participants provided written informed consent prior to enrollment.
Individuals with CLBP were considered for enrollment if they had a primary complaint of CLBP for at least 3 months prior to study enrollment, with an average daily pain intensity of ≥40/100 at worst on a 0 to 100 scale (0=“no pain,” 100=“worst pain imaginable”). Individuals were considered for enrollment as pain-free controls if they did not experience any LBP or chronic pain condition during the 3 months prior to study enrollment. Individuals either pain-free or with CLBP were excluded under the following criteria: (1) symptoms of lower extremity nerve root involvement, such as motor weakness and sensory disturbance; (2) prior surgery for low back musculoskeletal pain; (3) current use of opioids; (4) comorbidities, including uncontrolled hypertension, diabetic neuropathy, circulatory disorders interfering with activities of daily living, cardiac event history (eg, myocardial infarction), or epilepsy; (5) implanted cardiac pacemaker; (6) psychiatric-related hospital admission within the previous year; or (7) pregnancy. In addition, the onset of CLBP could not have resulted from trauma, such as a car accident, work accident, or fall. Finally, individuals with CLBP could not have received treatment for their pain condition by a health care professional within the previous month.
All prospective participants also were screened for cognitive impairment using the Mini-Mental State Examination (MMSE),30 with MMSE scores lower than 23 being excluded from the study. Lower MMSE scores were excluded to prevent enrolling of individuals who had cognitive decline that interfered with activities of daily living (ie, those with mild cognitive impairment [MCI] or greater) because (1) individuals with MCI may have difficulty reporting CLBP and accurately reporting psychological or cognitive factors and (2) cognitive influence on CLBP measures is most likely different for individuals with MCI compared with those without MCI. During cognitive screening, participants also were screened for intact hearing, which was a requirement for study testing.
Once enrolled, participants completed intake questionnaires for demographics, CLBP measures (daily pain, pain duration), and pain catastrophizing and underwent an assessment of verbal working memory. Participants also performed functional tasks related to the spine to assess movement-evoked pain intensity and pain-related disability.
CLBP Measures
Study dependent measures were daily pain intensity and movement-evoked pain intensity. Daily pain is an overall measure of pain intensity experienced in the previous day or week, and movement-evoked pain is a real-time measure of pain intensity specific to movement or function. Both measures are used in pain research, although movement-evoked pain is a more contemporary measure for which the applicability to chronic pain is accumulating.31,32 Moreover, both measures assess different aspects of the pain experience.
Daily pain intensity was assessed using the short-form Brief Pain Inventory (BPI). The BPI is a measure of a person's best and worst pain intensity over the previous 24 hours, as well as average and present pain intensity. Each rating was scored using an 11-point numeric scale, with 0 meaning “no pain” and 10 meaning “worst pain imaginable.” The 4 ratings were then averaged to arrive at a daily pain intensity score for each individual. The BPI has been deemed a valid and reliable measure of musculoskeletal pain.33
Movement-evoked pain intensity was assessed through functional tasks of the Back Performance Scale (BPS).34 The BPS determines how CLBP affects physical performance and consists of 5 physical tests requiring spinal movement to complete a particular task. This BPS is a reliable and valid measure of physical performance among individuals with low back pain.35,36 Participants rated their “pain rating at worst” during each of the tasks on a 0 to 100 numeric pain rating scale, with 0 meaning “no pain” and 100 meaning “worst pain imaginable.” The numeric pain rating scale has been deemed a reliable and valid measure of low back pain intensity and has been used extensively in low back pain clinical trials.37–39 A movement-evoked pain intensity composite score was created for each individual by averaging pain ratings for the 5 tasks, with higher scores indicating greater movement-evoked pain intensity. Use of a movement-evoked pain composite score was determined a priori and deemed appropriate because (1) intercorrelation of pain across the 5 tasks was between .60 and .85, (2) using an average of pain ratings rather than a single pain rating reduces the error variance, and (3) using a composite score increases the capacity for explained variance to examine associations.40
The BPS also was used to compare intake disability across groups. Each task was scored from 0 to 3, with 0 indicating the ability to perform the task without difficulty and 3 indicating the inability or unwillingness to perform the task (or maximum difficulty with the task). The cumulative BPS score for disability ranged from 0 (no disability) to 15 (maximum disability).
Working Memory and Pain Catastrophizing
Working memory was assessed via the digit span backward test, which is an established measure for assessing changes in verbal working memory during chronic pain.22 Moreover, a recent study showed the strongest association between digit span backward performance and chronic pain compared with other cognitive performance tests, including category fluency, story recall, and the doors test performance.23 The digit span backward test is a subtest of the Wechsler Adult Intelligence Scale (WAIS-IV) (Pearson, San Antonio, Texas). Participants read a series of numbers presented by a tester and repeated the number sequence backward. If successful, longer lists of numerical digits were read. A participant continued with the task until reaching 2 unsuccessful trials, at which time all successful trials were added to arrive at a raw digit span backward score. Both the raw score and the age-adjusted percentile are reported. Furthermore, each individual participated in a 3-digit span practice test prior to testing to familiarize him or her to testing and confirm comprehension of the digit span backward test.
Pain catastrophizing was assessed using the Pain Catastrophizing Scale (PCS).6 The PCS assesses exaggerated negative response as it relates to a painful experience, separated into constructs of rumination, magnification, and helplessness. The PCS has 13 items, and scores range from 0 to 52 (higher scores indicate higher catastrophizing). Multiple studies have determined pain catastrophizing to be a factor associated with CLBP.8,10,11
Data Analysis
Analyses were completed using IBM SPSS Statistics for Windows software, version 22 (IBM Corp, Armonk, New York). Alpha level was set at P=.05 for all analyses. Group (CLBP versus pain-free control) differences in continuous demographic and pain measures (age, pain duration, daily pain intensity, movement-evoked pain intensity, pain-related disability, working memory, and pain catastrophizing) were assessed using Mann-Whitney U nonparametric testing to account for unequal sample size and group variance. Chi-square tests assessed condition differences in categorical demographic factors (ie, sex, education, socioeconomic status).
Psychological and cognitive associations to CLBP measures (daily pain intensity, movement-evoked pain intensity) were examined with ordinary least squares (OLS) regression models. First, pain catastrophizing and working memory were entered separately to assess the unadjusted association with daily pain intensity and movement-evoked pain intensity. Next, pain catastrophizing and working memory were entered into multivariate hierarchical regression models to: (1) assess comparative associations between factors, (2) test for pain catastrophizing by working memory interaction, and (3) adjust for age, sex, and education. Separate multivariate models were created for daily pain intensity and movement-evoked pain intensity. Age (continuous), sex (male, female), and education (high school attendance, college attendance) were entered into the first block. Pain catastrophizing, working memory, and the interaction between pain catastrophizing and working memory were entered into the second, third, and fourth blocks, respectively. To control for multicollinearity, pain catastrophizing and working memory were entered as centered product terms, and interaction was a multiplication of these terms. Standardized regression coefficients (beta) assessed comparative strength of the factors in the final model block. One thousand-sample bootstrapping was used to calculate bias-corrected and accelerated 95% confidence intervals for the standardized regression coefficients.41 Absence of multicollinearity was confirmed via a priori cutoff rules for intercorrelation (r<.70), tolerance (>.20), and variance inflation (<4).42 Furthermore, a sample size of 10 participants per independent variable was determined a priori for all models to prevent overfitting and potential bias of estimates.43
Role of the Funding Source
This study was supported by funding from National Institutes of Health and National Center for Research Resources CTSA grants (R01AG039659, UL1 TR000064, and TL1TR000066) and the Department of Physical Therapy, University of Florida.
Results
Of the 217 individuals assessed for eligibility, 60 with CLBP and 30 who were pain-free were enrolled and analyzed. Groups were of similar age, proportion of female participants, education, and income (Tab. 1). Participants who were pain-free were validated based on very low daily pain and movement-evoked pain ratings that also were lower than the ratings for participants with CLBP (P<.001). Participants with CLBP had higher pain catastrophizing (U=386.50, P<.001) and lower working memory (U=1,206.50, P=.008) compared with participants who were pain-free.
Demographic and Pain Characteristics of Participants With CLBP and Control Group of Participants Who Were Pain-freea
As expected, no associations existed between CLBP measures and either pain catastrophizing or working memory for participants who were pain-free (P>.05). For daily pain among individuals with CLBP (Figure, graph A), working memory was inversely associated with daily pain intensity (R2=.14, P<.001), whereas pain catastrophizing was positively associated although not significant (R2=.06, P=.062). The final model accounted for 26.9% of the variance in daily pain (F6,59=3.26, P=.008). After accounting for age, sex, and education in participants with CLBP, working memory remained inversely associated with daily pain, accounting for approximately 7% variance (Tab. 2).
Pain catastrophizing and working memory associations with (A) daily pain and (B) movement-evoked pain (z-transformed). * Significant association (P<.05).
Adjusted Model of Daily Pain Associations Among Participants With Chronic Low Back Paina
For movement-evoked pain among individuals with CLBP (Figure, graph B), working memory was inversely associated (R2=.23, P<.001), whereas pain catastrophizing was positively associated (R2=.10, P=.016). The final movement-evoked pain model accounted for 27.8% of the variance (F6,59=3.64, P=.004). After controlling for age, sex, and education in participants with CLBP, only working memory remained an associative factor, accounting for approximately 14% of the variance (Tab. 3).
Adjusted Model of Movement-Evoked Pain Associations Among Participants With Chronic Low Back Paina
Discussion
Past CLBP research has not comprehensively examined working memory associations with established psychological measures. Here, working memory demonstrated the strongest association with daily CLBP and movement-evoked pain intensity compared with (and after accounting for) age, sex, education, and pain catastrophizing. This study advances understanding of working memory as it relates to CLBP and provides a framework for future prognostic CLBP research. Specifically, the extent to which working memory predicts CLBP, is associated once CLBP occurs, or moderates treatment effects can be a priority of future prospective studies.
Our findings corroborate previous work in which individuals with chronic pain had poorer working memory than pain-free controls.22 The observed inverse associations between working memory and pain also align with previous basic, preclinical, and clinical research.26 Collectively, the evidence implies that impaired working memory is a consequence of pain. Moriarty et al26 proposed a preliminary model whereby neuroplasticity, neuromediators, or limited cortical resource mechanisms explained how pain induced cognitive impairment. While plausible, the current stance is based on a unidirectional relationship between pain and cognitive performance. However, working memory interacts with multiple brain systems to coordinate attention, long-term memory, and function, which are processes necessary for the perception of musculoskeletal pain.18,19,44 Therefore, impaired working memory may be more than a consequence of CLBP by also contributing to elevated pain perception. The magnitude of association observed with movement-evoked pain—a dynamic, real-time pain measure—lends credence to this hypothesis. Although untestable in the current study design, these findings align with a recent investigation of patients who had undergone surgery whereby impaired cognitive performance was found to predict pain chronicity at 6 and 12 months.45 Future longitudinal studies in nonsurgical cohorts, specifically those with acute LBP, will be able to assess working memory as a predictor of LBP chronicity.
Recent neurocognitive theories support the notion of impaired cognitive performance influencing higher movement-evoked pain. Legrain et al46 proposed working memory to operate through top-down modulation by directing attention toward goal-specific stimuli while inhibiting attention toward goal-irrelevant stimuli. Conceptually, individuals with poorer working memory performance, or perhaps higher working memory load, may be unable to attenuate pain-related attention during a functional task, thereby elevating the pain (or salience of pain) perceived.47,48 Work in this area is both equivocal (higher cognitive load has been found to attenuate experimentally induced pain,49) and under-researched (a paucity of studies exists on the interface among working memory, attention, and clinical pain conditions).
Recently, Moseley and Vlaeyen50 proposed that imprecise cortical encoding of events leads to associations between nociceptive and nonnociceptive input and that the magnitude of this imprecision influences pain generalization to similar, previously nonpainful events. In other words, repeated association of movement and pain can transform both related and unrelated movements into conditioning stimuli that invoke a painful response. Although imprecise encoding may not be a product of working memory,50 the resultant pain generalization requires working memory to coordinate present attention to movement with past painful experiences. Future prospective studies will examine: (1) how working memory changes from acute to chronic pain, (2) how working memory changes mediate or moderate movement-evoked pain intensity, and (3) the extent to which working memory and movement-evoked pain contribute to chronic pain neuroplasticity.
Absence of pain catastrophizing associations was a key finding and may be the product of pain outcome measures. Historically, pain catastrophizing has been assessed in conjunction with current resting pain intensity but not with movement-evoked pain intensity.8,11,51 Resting pain intensity likely involves more affective processes because individuals cannot directly attribute the pain to a movement or source. Resting pain intensity also can occur without attention to other tasks, which may increase the salience of the pain experienced and the distress toward pain consequences. In contrast, movement-evoked pain intensity is assessed shortly after task performance, which may not be adequate time to emotionally process pain consequences. Moreover, movement is the perceived source of pain, so catastrophizing over the implications of an already completed task is less likely than catastrophizing over the potential for future pain-related disability and consequent avoidance of tasks from such thoughts. Pain catastrophizing has demonstrated association with daily pain intensity, as well,9,10 although not in this study. Nevertheless, the prognostic value of pain catastrophizing is established, and the findings here only downplay associations with movement-evoked pain intensity (and potentially daily pain intensity) once chronicity has occurred.
Age specificity also may explain the lack of pain catastrophizing associations. Picavet et al11 found pain catastrophizing to predict CLBP 6 months after initial assessment. However, although the study examined more than 1,500 participants, more than 50% of their sample comprised middle-aged adults. Moreover, this study excluded individuals aged 65 years and older. Similarly, Smeets et al8 found that reducing pain catastrophizing mediated physical and cognitive-behavioral treatment responses (ie, pain intensity and disability reduction) among individuals with CLBP. However, the average age of the participants was 43 years, and individuals over 65 years were excluded. In contrast, our study used equal representation across the life span. Riley et al52 performed an age group comparison of psychological distress and chronic pain and found the strongest associations in the middle-aged cohort. This phenomenon was proposed to be related to the onset of chronic pain during middle age, coupled with increased life responsibilities.52 The same study demonstrated the weakest associations between psychological distress and chronic pain among older adults,52 which may be related to a reduction in life responsibilities (eg, occupation, family support). Age group comparisons were not the intent of this study; however, decreased association between pain catastrophizing and pain in a cohort that is more representative of the life span supports future investigation into age-specific CLBP mechanisms.
There are limitations to consider when assessing the study findings. First, our analysis was restricted to a single session. Determining the extent to which poor working memory performance predicts LBP chronicity, as established with high pain catastrophizing, is the aim of future studies. Second, we did not use a comprehensive assessment of working memory, although verbal is the most researched working memory construct.19 Moreover, the digit span backward test is an established measure of working memory in pain research, and results were in line with previous findings.22,23 Third, we did not account for sleep, which could potentially explain working memory differences between individuals with CLBP and those who are pain-free. However, we did account for age and education, which are known to affect working memory, and excluded opioid use, which is not always considered in cognitive studies.26 Finally, we did not assess moderation by other cognitive or psychological factors, such as attention23 or depression,53 because the primary objective was to compare differential association of working memory and pain catastrophizing on CLBP. These factors and others (eg, affect, resilience, self-efficacy) should be considered in conjunction with working memory and chronic pain.
There are also multiple study strengths. First, we compared working memory associations between individuals with CLBP and those who were pain-free. A recent systematic review by Moriarty et al26 identified lack of comparisons between individuals with pain and those who were pain-free as a limitation for studies assessing cognitive influences on chronic pain. Second, only individuals with CLBP were included in the study. Previous studies have utilized heterogeneous pain populations where CLBP prevalence was low or not reported.52,54 Studies by George and colleagues55,56 showed pain-related fear and depression to have similar associations across musculoskeletal conditions, so location may play a limited role. However, a study by Apkarian et al57 showed impaired cognitive performance in individuals with chronic back pain but not in those with neuropathic pain. Third, both cognitive and psychological factors were assessed in the same CLBP model, which is not routinely performed. Finally, we assessed both a recall measure (daily pain) and a real-time functional measure (movement-evoked pain) of CLBP intensity. These assessments allowed us to examine a differential association based on type of pain measure and provided a more comprehensive look at each individual's pain experience.
In summary, working memory was associated with daily and movement-evoked CLBP above and beyond established CLBP factors such as pain catastrophizing. Future research will confirm whether working memory is a consequence or predictor of CLBP, as well as the extent to which working memory influences real-time pain intensity. Whether the association between pain catastrophizing or working memory and CLBP depends on age is also a priority of future research.
Footnotes
Dr Simon, Dr Bishop, Dr Riley, Dr Fillingim, and Dr George provided concept/idea/research design. Dr Simon, Dr Bishop, Dr Riley, and Dr George provided writing. Dr Simon and Mr Lentz provided data collection. Dr Simon, Dr Bishop, and Dr Fillingim provided data analysis. Dr Simon provided project management and participants. Dr Simon, Dr Riley, and Dr George provided fund procurement. Mr Lentz, Dr Riley, Dr Fillingim, and Dr George provided consultation (including review of manuscript before submission). The authors thank Kim Sibille, PhD, and Aliyah Snyder, BS, for their assistance in interpreting study findings.
This study was approved by the University of Florida Institutional Review Board.
This study was supported by funding from National Institutes of Health and National Center for Research Resources CTSA grants (R01AG039659, UL1 TR000064, and TL1TR000066) and the Department of Physical Therapy, University of Florida.
- Received June 10, 2015.
- Accepted December 13, 2015.
- © 2016 American Physical Therapy Association
References
Issue highlights
