Imaging Biomarkers for Myofascial Pain

Definition of proposed composite multimodal biomarker-The investigators propose to develop a quantitative tissue-level classifier based on quantitative metrics (features) derived from ultrasound elastography, Doppler, bioimpedance spectroscopy and high-density electromyography, as an indicator of the normal biological process in myofascial tissues, and pathogenic process in active and latent phases of myofascial pain.

Overall approach and scientific rigor: In Aim 1, the investigators will develop methods to generate reproducible metrics (features) from the raw tissue-level measures and determine the minimum detectable change in these features in a pilot study. In Aim 2, the investigators will conduct a longitudinal observational study with two groups of subjects (control and myofascial pain). The investigators will develop a classification algorithm that optimally differentiates between active and latent phase of myofascial pain and normal myofascial tissue.

Study population and anatomical site. The chosen pain condition is chronic neck and shoulder pain. The investigators will recruit two groups of subjects: Group 1: Chronic myofascial pain as determined by baseline clinical examination using Travell and Simon's criteria7 and Group 2: pain free controls. The investigators will focus on two standardized anatomical locations (Figure 4). This will enable imaging the medial upper trapezius and the infraspinatus muscles, which are common locations for MTrPs55 as well as the levator scapulae. These three muscles have quite different morphology and fasciae45. The levator is a fusiform muscle with well-defined fascia that includes the muscle while the trapezius has thinner fascia from where perimysium septae cross the muscle belly. The infraspinatus has multiple fascial layers on its surface and has clear segmental linkages to the C5-6 segment56 Eligibility criteria: The investigators will recruit adults 18-65 years of age. Exclusion criteria: (1) diagnosis of fibromyalgia, chronic fatigue syndrome or chronic Lyme disease; (2) Diagnosis of cervical radiculopathy, neuropathy, or neuritis; (3) History of head, neck, cervical spine, or shoulder girdle surgery; (4) Atypical facial neuralgia; (5) New medication or change in medication in past 6 weeks; (6) Current throat or ear infection.

Masking and Matching: This is a single-blind longitudinal observational study. The team performing the data collection and analysis will not know the group allocation of the subjects and will be blinded to the results of the clinical evaluations. The two groups will be age and gender matched using a paired recruitment strategy57. The investigators will identify a pool of eligible control subjects with no history of pain and divide them into gender and age brackets (18-30; 31-50; and >50). For each Group 1 subject recruited in a bracket, the investigators will recruit a matched Group 2 subject from the pool.

Sex as a biological variable: Myofascial pain is widely prevalent in the community and affects both men and women. Trapezius myalgia is more prevalent in women58. The investigators will utilize age and gender-matched groups, and will test the classifier performance for both the pooled population as well as separately by gender to identify any gender-specific differences in the biomarker measures.

Outcome Measures: The primary outcome measure will be the composite classifier based on the tissue-level quantitative biomarkers. The investigators will perform repeated data collections every month for 3 months. The clinical phenotype of the subjects (normal, latent, episodic active, and persistent active) will be determined by a comprehensive physical examination protocol12. The investigators will utilize the NIH HEAL Common Data Elements for adult chronic pain to collect self reports. To further characterize the clinical phenotype, as a secondary outcome measure, the investigators will utilize an ecological momentary assessment (EMA) application (Metricwire) on a smartphone to obtain a daily pain rating triggered at random points during the day and collect automated activity monitoring from the smartphone sensors. The investigators will also collect weekly 3-item pain intensity and interference59.

Data collection procedures Data management: This is a single site study. All study procedures will be performed at Mason. The study biostatistician (Rosenberger) will set up the appropriate masking controls and electronic case report forms (eCRFs) in the electronic data capture system (REDCap). All study data will be entered into REDCap using eCRFs. Study personnel will have appropriate role-based access controls in REDCap. Source validation will be performed using REDCap's built-in checks.

Masking: A single clinician (Gerber) will obtain each subject's consent and conduct history and the physical examination. An additional clinician (DeStefano) may be present to assist, and a research assistant will be present to take notes and enter data. The engineering team, supervised by the PI (Sikdar) and co-I Chitnis, will collect the outcome measures in a separate room and will be masked to the patient's history and results of the physical examination. A manual of operating procedures will be developed for the study.

Data analysis procedures. All data analysis will be performed by a biostatistics graduate research assistant under the supervision of the data scientist (Lee) and study biostatistician (Rosenberger).

Primary analysis: The investigators will construct and rigorously validate a multi-class classification algorithm based on functional time series and statistical learning methods. Here, the biomarker time series can be represented as combination of unique temporal patterns/signals, or basis functions. These functions include time-invariant eigenbasis functions80, smoothing splines81, wavelets82, or functional principal components83. Using functional data analysis, a composite predictor variable will be constructed that summarizes the pertinent information contained in the biomarker time series. Then, a multi-class classification method will be constructed using supervised learning approaches, such as support vector machines84, discriminant analysis85,86, neural networks87,88, regression trees89. The classifying algorithm will use the composite predictor to codify subjects into the four relevant categories (pain - episodic, pain - active, control-episodic, and control-active). K-fold cross-validation will be used to assess the classifier's accuracy based on sensitivity, specificity, F1 score, and the area under the ROC curve for multi-class scenarios90,91.

Secondary analysis: Several secondary analyses will be performed including: (1) Determine normative values of biomarkers in control group (Group 2); (2) Evaluate convergent validity of primary and secondary biomarkers. Since the underlying ground truth cannot be measured directly, the primary and secondary biomarkers will be utilized to evaluate convergent validity; (3) Correlation with corresponding clinical measure (range of motion, pressure pain threshold.