Multidimensional analysis of the clinical spectrum and symptom burden of unexplained myofascial pain.

Out of 137 subjects screened, 96 subjects met eligibility criteria and were recruited for the study. Complete records for all variables were available for 82 subjects. Reasons for the 14 missing records were the following: (1) Did not complete multiple questionnaires (N= 5); (2) Did not answer all fields on one questionnaire (N=5; PCS: 2 incomplete records; PEG: 2 incomplete records; PHQ-2: one incomplete record); (3) Missing age (N=2); (4) Missing PPT (N=2). Subjects with complete data records for all selected variables (N=82) were included in this analysis. Descriptive characteristics (mean ± standard deviation or percentage) of the variables (physical findings and self-reports) in the study sample are shown in Table 1 , separately for the active, latent and normal groups and for all subjects. Current bilateral neck/shoulder pain was reported by 41.25% of the participants, while 17.5% reported unilateral pain and 41.25% of participants did not report experiencing pain currently. In our sample, 36.4% reported both neck and shoulder pain, 14.3% reported neck pain only, 11.7% reported shoulder pain only, while 36.6% reported neither neck nor shoulder pain. Physical findings of MTrPs and taut bands were ubiquitous in the sample: bilateral findings were present in 79.3% of the subjects, unilateral findings were present in 11%, and there were no findings in only 9.7%. Based on the physical exam, 28.1% were determined to have active myofascial pain, 46.3% were determined to have latent myofascial pain, and the remaining 25.6% had no myofascial pain. Physical findings in normal subjects were not considered to meet Travell and Simon’s criteria for myofascial involvement. Individual measures exhibit large variability and overlap among subjects in the active , latent and normal groups ( Figure 2 ). In particular, there is strong overlap on PEG between the latent myofascial pain and normal groups, demonstrating that subjects determined to be asymptomatic by myofascial criteria may have symptom burden similar to those presenting with latent MP. Based on self-reported outcomes, the normal group was not truly asymptomatic. Table 2 shows the regression analysis for PEG as a dependent variable. In the pooled sample, 75% of the variance is explained by the predictors and adjustment variables. Partial R 2 values represent the proportion of variance explained by each predictor after adjusting for other variables. Physical Function and PCS are the most important predictors in the model, with partial R 2 of 25.6% (p<0.0001) and 16.7% (p=0.0004), respectively. As expected, the adjustment variable, pain group, is important, so we examine the subgroup analysis defined by the group ( Tables 3 , 4 and 5 ). The predictors and adjustment variables explain 85% and 92% of the variance in the active and normal groups but only 46% of the variance in the latent group. In the active group, the most important predictor is PCS, with a partial R 2 of 37.5% (p<0.02). In the normal group, Physical Function (70.7%, p=0.0006), PCS (41.9%, p=0.023), Hypermobility (35.6%, p<0.05), and GAD-2 (34.4%, p<0.05) all show substantial partial R 2 values. In the latent group, PHQ-2 is the most important predictor, with a partial R 2 of 20.2% (p=0.014). Sex at birth is an important adjustment variable in both the active and normal groups, but not in the latent group. Variance inflation factors indicated no evidence of substantial multicollinearity across the three regression models. For the active and latent groups, pairwise partial correlations with graphical LASSO resulted in models with a median of 14 (interquartile range 12–15) and 4 (interquartile range 3–6) non-zero edges, respectively ( Figure 3 ). No non-zero edges could be reliably found in the normal group. In Figure 3 , for the active group, physical function is negatively correlated with hypermobility (−0.43) and PEG (−0.37). PCS is positively correlated with both PEG (0.38) and GAD-2 (0.36). In contrast the latent group exhibits a different network topology, with a negative correlation between Physical Function and PCS (−0.26), which is weakly correlated with PEG (0.06). PHQ-2 and GAD-2 remain positively correlated (0.31), while the negative correlation between PPT and QST is weak (−0.07). See Supplement for additional details.

This cross-sectional study was conducted at George Mason University as part of a larger study to develop biomarkers for MP ( clinicaltrials.gov NCT#: NCT06060925 ). For differentiating between two clinically-relevant subgroups with an area under the receiver operating curve of 0.75, we determined that at least 19 subjects in each subgroup are needed to obtain a power of 0.8 and significance of 0.05. Participants were recruited from the general community, and a variety of different healthcare private practices in the greater Virginia/Washington DC/Baltimore area, from January, 2023 to December, 2024. All study procedures were approved by our Institutional Review Board and informed consent was obtained from all participants. Figure 1 shows the flowchart for participant recruitment and assessment. We recruited adult participants (≥18 years) with persistent or episodic pain in the neck/shoulder region for at minimum the past 3 months, as well as individuals without pain (no self-reported persistent or episodic neck or shoulder pain for the past 5 years). Participants who reported the following during phone screening were excluded: 1) previous diagnosis of fibromyalgia, chronic fatigue syndrome or chronic Lyme disease; 2) previous diagnosis of cervical radiculopathy, neuropathy or neuritis; 3) history of recent head, neck or shoulder girdle surgery; 4) new medication or change in medication in past 6 months; 5) history of shoulder or cervical fracture. For participants meeting eligibility criteria, an in-person comprehensive history and musculoskeletal physical exam was performed by experienced clinicians (two board-certified physical medicine and rehabilitation physicians and two licensed physical therapists). The standardized physical examination 9 included identification of active and latent MTrPs in the neck and shoulder region following Travell and Simon’s criteria 27 . Clinicians noted findings including the presence of MTrPs, “knots”, taut bands, asymmetry of palpation compared to contralateral side, the presence of a motor abnormality (physical finding in the muscle tissue) and a sensory abnormality (tenderness and referred symptomatic pain) 7 , 8 . Based on this in-person physical examination, subjects with neck/shoulder pain but without myofascial involvement were excluded ( Figure 1 ). Based on eligibility screening and physical examination, subjects were classified as either active (spontaneous persistent or episodic myofascial pain without provocation in the neck/shoulder region ≥3 months), latent (myofascial pain in the neck/shoulder region that is reproduced only upon provocation/palpation), or normal (clinically asymptomatic; no self-reported chronic pain in past 5 years; no palpable findings that reproduce pain in the neck/shoulder region; may have asymptomatic palpable findings). Determination of MP was therefore made at the whole-person level, i.e., active , latent or normal subject, rather than at the regional level, i.e., presence of an active or latent MTrP. This was done to allow analysis using other whole-person measures. Details on outcome measures are provided in the Appendix . We utilized the following standardized adult chronic pain NIH HEAL CDE 26 : Pain, Enjoyment of Life and General activity score (PEG) 28 ; Pain Catastrophizing Scale (PCS) 29 ; PROMIS Physical Function 30 ; PROMIS Sleep Disturbance Scale 31 ; Generalized Anxiety Disorder Scale 2-item scale (GAD-2) 32 ; Patient Health Questionnaire 2 (PHQ-2) 32 . Beighton score 34 modified with Brighton 35 to assess for hypermobility syndrome. Pressure pain threshold (PPT) 36 , 37 and quantitative sensory testing (QST) for windup 38 . Regression analysis was designed to evaluate whether the symptom burden, measured using the HEAL CDE 26 , along with validated measures of hypermobility and sensitization, can explain the variability in the clinical spectrum PEG in MP. Network analysis was used to investigate partial correlations among the different measures. Only subjects with complete data records for all selected variables were included in this analysis. Summary statistics are presented in Table 1 and in Figure 2 . Linear regression analyses were conducted with PEG as the dependent variable. Predictors were GAD-2, hypermobility, PCS, PHQ-2, physical function, sleep disturbance, PPT, and QST. The model was adjusted for sex at birth (male/female), age (considered continuous), and pain groups ( active, latent and normal ). Model aptness was verified using residual plots, and multicollinearity was assessed by analysis of variance inflation factors. Subgroup analyses were performed as warranted. To explore the relationships among variables, we applied graphical LASSO to construct partial correlation networks 39 , 40 . The analysis was performed using the qgraph package in R (version 4.3.3). For the penalty term λ, we first determined the upper bound as the smallest value of λ that results in a network with no connected edges. The minimum value of λ was set to 0.01 times the upper bound, ensuring a fully connected and sufficiently complex network. A sequence of equally-spaced λ values was then generated within this range, and the optimal λ was selected by minimizing the Extended Bayesian Information Criterion (eBIC) 25 .

Self-reports and objective measures highlight a complex biopsychosocial burden of symptoms in the patient with MP and explains the observed variability in pain intensity and interference. Network analysis can elucidate interconnections and mediation among multidimensional factors and provides an innovative whole-person clinical approach for risk stratification driven by symptom burden that does not solely rely on identifying specific local pain contributors. Our findings can inform future longitudinal studies aimed at elucidating the directionality, temporality and predictive value of biopsychosocial contributors to the clinical spectrum of chronic non-specific cMSKP with myofascial involvement.

This study presents a clinically relevant, multidimensional approach to evaluating the burden of symptoms associated with MP, in the absence of a clear consensus on how MP should be assessed or managed 8 , 17 . Our findings build upon previous literature in three important ways. First, our data revealed significant variability in the self-reports and objective findings across clinically-relevant subgroups in a cohort of subjects recruited from the general community. Second, while the variability was not unexpected since our eligibility criteria were intentionally broad, we found that a multidimensional analysis using assessments across physical and psychosocial domains explained much of the variability in pain intensity and intereference. Third, a network analysis identified linkages among clinically relevant variables (nodes) explaining potential pathways contributing to identifying which symptoms are prevalent and their variability. The active group demonstrated a high symptom burden with links between higher PEG, poorer physical function, hypermobility, higher levels of pain catastrophizing (PCS), and higher levels of anxiety and depression (GAD and PHQ-2); and higher levels of sensitization based on PPT and QST, linked to sleep disturbance. This is not the case for the latent group, which had fewer links (PHQ-GAD; and PCS-PF). In normal subjects, the graphical LASSO yielded a fully sparse precision matrix, with no edges retained at the selected regularization level with insufficient evidence for strong conditional associations among the variables. These collective findings provide novel early evidence supporting the need to understand how physical, emotional, social, and environmental factors all shape how symptom burden is individually experienced, and the need for prioritizing targeted interventions to improve whole-person outcomes. Our findings in MP are consistent with previous literature on chronic pain networks 18 – 22 . Pain catastrophizing 41 was shown to strengthen the association between pain and both anxiety and depression 18 . The limbic system mediates pain by integrating emotional and motivational dimensions with inputs from the somatosensory cortex 42 – 45 and interacts bidirectionally with the hypothalamus-pituitary-adrenal (HPA) axis 46 . Thus, chronic stress may lead to HPA dysregulation and pain sensitivity creating a cycle of chronic pain and stress 47 . Notably, patients with chronic upper trapezius MP demonstrate increased limbic system (i.e., anterior insula) activity 45 . Our results do not imply causal links between physical and psychosocial components and the psychosocial burden may result from the stress of navigating the healthcare system when faced with the need for managing multiple symptoms 48 , which in turn can perpetuate pain catastrophizing. One novel finding in the network structure is the link between hypermobility and physical function, which shows a strong partial correlation in the active group but not in the latent group. A higher prevalence of hypermobility in MP has been reported 49 , and our results indicate that physical function mediates the association between hypermobility and pain. While active and latent MTrPs are currently used for classifying the physical findings associated with MP, evidence supports the hypothesis that the regional findings in MP are linked to multivariate central mechanisms that may be common across many somatic, neurologic, and visceral pain conditions 12 , 45 . The “myofascial unit”, defined as an integrated anatomical and functional structure that includes muscle fibers, fascia (endomysium, perimysium and epimysium) and its associated innervations (free nerve endings, muscle spindles), lymphatics, and blood vessels 23 , 24 , 50 , 51 , has been implicated to play a foundational role 8 , 9 , 12 , 52 – 61 . Experimental data help link mechanisms of central sensitization, a common component of all chronic pain syndromes, to dysfunction at the myofascial unit through the mechanism of neurogenic inflammation 58 , 59 , 62 , 63 . For example, several studies have reported soft tissue findings associated with pain conditions of visceral hypersensitivity such as irritable bowel syndrome and chronic pelvic pain 64 – 66 , which raises the possibility that a primary pathology residing within somatic or visceral tissues 59 , such as facet joints, irritable bowel, and endometriosis, etc. could contribute to dysfunction of the myofascial unit. A critical impediment to clinical and mechanistic research in MP is the lack of reliable and reproducible diagnostic criteria 10 , 11 , 13 , 17 . While imaging and other biomarkers could improve reproducibility 54 , the clinical variables we selected for this study 26 , explained 75% of the variance of PEG in our sample of patients with MP. In contrast, the presence of physical findings was ubiquitous ( Table 1 ). Our findings provide evidence for a shift away from relying solely on the unreliable identification of MTrPs and other regional pain contributors towards assessments that might provide insight into managing the symptom burden of the whole person. Future research needs to investigate individual networks of causal linkages from longitudinal observations of relevant biopsychosocial variables to understand individual differences in pain perception across time 67 , 68 , including in subclinical latent phenotypes, and may lower the number of patients labeled non-specific cMSKP 3 . The links between the central and peripheral nervous system and the myofascial unit need further exploration. Traditional clinical practice in MP has been siloed by regional symptomatology, with an emphasis on MTrPs as the target for assessment and treatment. This approach fails to adequately capture the complex reciprocal systemic biopsychosocial interactions, making it challenging to identify which treatments are likely to work for which subject. While psychosocial symptoms are indicative of poor outcomes 69 – 74 , current predictive and prognostic models 75 , 76 do not provide insight into how to alter outcomes. Network models might provide a new approach to meaningfully integrate information across biopsychosocial domains to guide mechanism-informed clinical decision-making 19 , 20 . Key central influencers within individual networks can offer targets for multidisciplinary therapeutic interventions 77 , 78 . It is our view that the larger symptom burden of the active clinical profile is likely to require earlier, and more intense and comprehensive care and may experience a longer recovery time compared to the latent profile which exhibited lower symptom burden. Identifying patients with higher risk profiles can avoid unnecessary testing and interventions for lower risk individuals. Potential sources of bias include incomplete records from 15% of recruited subjects, and more women in the active group. While the overall sample size is adequate for regression, the sample size of the active group was small for a multivariate network analysis. These findings need to be confirmed in a larger sample. The network stability was moderate for this sample size although the network structure was not sensitive to parameter choice (see Supplementary material ). Our analyses were exploratory without adjustment for multiple testing, so p-values should serve as a guide rather than a formal level of significance.

Chronic musculoskeletal pain (cMSKP) is a leading cause of disability worldwide. Most patients lack a clearly identifiable underlying cause of their symptoms 3 . The traditional approach to studying cMSKP focuses on regional anatomy (e.g., low back, neck, shoulder, knee pain). Myofascial tissues are commonly involved in cMSKP, and are often a treatment target 4 – 8 . There is no definitive test to diagnose myofascial pain (MP), which is characterized by a motor abnormality (physical findings within the muscle) and a sensory abnormality (tenderness and referred pain) 7 , 8 . Diagnosis is commonly based on clinical assessments, including pain regionality, presence of palpable, symptomatic myofascial trigger points (MTrPs) in affected tissues, history, and physical examination 9 . Strong digital pressure on symptomatic MTrPs often exacerbates the patient’s spontaneous pain complaint and mimics the patient’s familiar pain experience. This approach has low to moderate reproducibility and reliability 10 , 11 . The symptom burden exhibits large variability between individuals and over time, and the presence or absence of local MTrPs does not fully capture the complexity of the patient’s lived daily experience. It remains unclear whether MP is a distinct regional syndrome, or part of the clinical spectrum (and component) of non-specific cMSKP 12 , 13 . Evidence suggests that the clinical manifestations, lived experience and pain-related functional limitations of cMSKP (and chronic MP) are strongly influenced by physiological, functional and psychosocial determinants 14 , 15 . The prevailing focus on regional contributing factors fails to address these complex biopsychosocial interconnections, and has led to overly simplified treatment approaches with limited effectiveness for patient outcomes 16 . Inconsistent nomenclature 17 further adds to the challenges for clinicians, patients and researchers. A growing body of research points to the need for a paradigm shift towards considering chronic pain as an emergent property of a complex system with central and peripheral components 18 – 24 . Network analysis is an emerging approach to studying complex systems by identifying system components (nodes) and their relationships (edges) 25 . However, to our knowledge, no previous study has 1) identified whether multidomain biopsychosocial determinants can explain the clinical spectrum of MP; nor 2) whether the network dependencies among these outcomes differs between the clinically defined phenotypes of MP (active, latent) versus normal. To address this gap, our primary objective was to investigate whether the symptom burden, measured using a set of patient reported outcomes from a standardized chronic pain common data elements inventory 26 , along with validated measures of hypermobility and sensitization, can explain the variability in the clinical spectrum of pain intensity and interference in MP. Our secondary objective was to determine whether the network dependencies differ between active MP, latent MP and normal clinical phenotypes.