Neurophysiological and transcriptomic Predictors of Chronic low back pain: study protocol for a longitudinal inception cohort study

Abstract

Chronic low back pain is one of the most common, costly, and debilitating pain conditions worldwide. Increased mechanistic understanding of the transition from acute to chronic low back and identification of predictive biomarkers could enhance the clinical assessment performed by health care providers and enable the development of targeted treatment to prevent and/or better manage chronic low back pain. This research protocol was designed to identify the neurological and transcriptomic biomarkers predictive of chronic low back pain at low back pain onset. This is a prospective descriptive longitudinal inception cohort study that will follow 340 individuals with acute low back pain and 40 healthy controls over 2 years. To analyze the neurophysiological and transcriptomic biomarkers of low back pain, the protocol includes psychological and pain-related survey data that will be collected beginning within six weeks of low back pain onset (baseline, 6, 12, 24, 52 weeks, 2 years) and remotely at five additional time points (8, 10, 16, 20 weeks, and 18 months). Quantitative sensory testing and collection of blood samples for RNA sequencing will occur during the six in-person visits. The study results will describe variations in the neurophysiological and transcriptomic profiles of healthy pain-free controls and individuals with low back pain who either recover to pain-free status or develop chronic low back pain.

1. INTRODUCTION

Low back pain (LBP) is the leading cause of years lived with disability worldwide (Wu et al., 2020). As the prevalence of LBP related disability has risen over the past two decades, so has the cost and use of health care services, with low back and neck pain accounting for $134.5 billion in direct costs annually (Dieleman et al., 2020). Axial LBP is the most common type of LBP and the pain is confined to the lumbar spine region. Although current guidelines for axial LBP recommend a conservative and non-surgical treatment approach, the cost of care and rate of long-term disability have continued to rise (Cruz et al., 2020; Kim et al., 2019; Zaina et al., 2020).

1.1. Factors involved in the transition from acute to chronic low back pain

Worldwide, it is estimated that up to 88% of individuals will suffer an acute low back pain (aLBP) episode of less than 6 weeks duration at some point during their lifetime (Hartvigsen et al., 2018). Pathological abnormalities such as disk herniation and lumbar stenosis are sometimes implicated as a source of LBP and are treated according to the etiology. However, 90% of people with axial LBP do not have an identifiable structural pathology or underlying disease (Maher, Underwood, & Buchbinder, 2017). Of the people presenting with axial aLBP, 39% will transition to chronic LBP (cLBP), pain lasting > 3 months of the last 6 months. Despite standard treatment, over 35% of individuals who develop cLBP will continue to report variable levels of disability for 12 months or longer (Foster et al., 2018). Currently, the mechanisms by which aLBP resolves for some individuals or transitions to cLBP for others remains unknown. Identification of the mechanisms involved in the transition from aLBP to cLBP could enable the development of targeted treatment to prevent or better manage cLBP.

Our team and others have identified several demographic, psychological, and environmental risk factors which are predictive of cLBP, including socioeconomic status (Shmagel, Foley, & Ibrahim, 2016; Starkweather, Ramesh et al., 2016), body mass index (Chowdhury et al., 2014; Su et al., 2018), depression (Bener et al., 2013; Carroll, Cassidy, & Côté, 2004), pain catastrophizing (Meints et al., 2019; Picavet, Vlaeyen, & Schouten, 2002), and a previous aLBP episode (Hestbaek, Leboeuf-Yde, & Manniche, 2003). However, these risk factors perform poorly when applied to clinical populations and are not precise enough to explain the physiological mechanisms of cLBP. In contrast, neurophysiological and biomarker predictors of cLBP are understudied despite increasing evidence that cLBP is associated with somatosensory abnormalities (Starkweather, Lyon et al., 2016) as well as transcriptomic changes within the immune (Lim et al., 2020; Teodorczyk-Injeyan, Triano, & Ineyan, 2019), peripheral (Hartlehnert et al., 2017; Parisien et al., 2019) and central nervous system (Yu et al., 2014).

1.2. Neurophysiological factors involved in the transition to chronic low back pain

Quantitative Sensory Testing (QST) is a non-invasive examination of pain perception and somatosensory function that has been used to detect changes in a variety of pain conditions, including LBP (Rolke et al., 2006; Starkweather, Heineman et al., 2016). Through the application of thermal, mechanical, and electrical stimuli at controlled intensities, QST can detect abnormalities in large, A-beta and A-delta, and small, C-sensory fibers and in the pain modulatory system (Edwards et al., 2016). Specific to LBP, studies using QST have found that cLBP patients display elevated peripheral and central sensitization (Giesecke et al., 2004; O’Neill et al., 2007).

Peripheral sensitization occurs when inflammatory mediators, released in response to damaged tissue, increase nociceptor excitability and decrease the threshold potential of A-delta and C primary afferent nerve fibers resulting in hyperalgesia localized to the site of injury (Hucho & Levine, 2007). In patients with LBP, peripheral sensitization presents as decreased mechanical and thermal pain threshold at the lumbar region (Aoyagi et al., 2019; Latremoliere & Woolf, 2009). Conversely, central sensitization refers to the amplification of neuronal signaling within the central nervous system (den Bandt et al., 2019; Pelletier, Higgins, & Bourbonnais, 2015; Woolf & Salter, 2000). Studies using QST have reported that patients with cLBP display symptoms of central sensitization including widespread hyperalgesia and allodynia, decreased tactile acuity, and spontaneous pain (Correa et al., 2015; O’Neill et al., 2007). Conditioned pain modulation (CPM) is an experimental paradigm that can be used to measure functioning of the endogenous pain inhibitory pathways. There is some evidence that impaired CPM may contribute to heightened risk for transition from acute to cLBP (Mlekusch et al., 2016).

While these findings are promising, few studies have compared the neurosensory profiles of individuals over the transition from acute to cLBP and heterogeneity in study design, testing measures, and sample size limit the conclusions. Further research using QST could allow for the identification of an individualized neurosensory profile to aid in clinical assessment and enable precise treatment plans targeting the underlying mechanisms of cLBP.

1.3. Transcriptomics and the transition to chronic low back pain

Transcriptomics encompasses the study of ribonucleic acid (RNA) transcripts that are produced by the genome at a given time (Stark et al., 2019). As opposed to deoxyribonucleic acid (DNA), RNA is dynamic with synthesis and degradation regulated by the internal and external environment. RNA-sequencing enables researchers to measure which genes are active in a cell and the level of transcription, identify biomarkers differently expressed between the health and diseased state, and explore the molecular mechanisms underlying a phenotype, in this case cLBP (Starobova et al., 2018). This approach has been used successfully to identify differentially expressed genes associated with a variety of chronic pain conditions including trigeminal neuralgia (Flegel et al., 2015), chemotherapy induced peripheral neuropathy (Page et al., 2019), and adhesive capsulitis (Gatchel et al., 2018). More recent work has shown that increased expression of genes located within the major histocompatibility complex (MHC) locus at LBP onset significantly influences the risk of cLBP (Dorsey et al., 2019). Therefore, longitudinal studies, beginning at pain onset, are essential to gain mechanistic understanding of pain chronicity, and in the identification of pain trajectories and predictive pain biomarkers (Croft et al., 2015).

1.4. Study overview

Herein, we describe a prospective longitudinal research protocol which is following 340 participants recruited at LBP onset and 40 healthy controls for two years. We aim to identify a comprehensive transcriptomic signature of cLBP and identify neurological and transcriptomic biomarkers which can predict cLBP at LBP onset. The specific aims of the study are to: (1) Characterize the physiological (somatosensory function), psychological (coping, reactivity, mood, stress), clinical (pain intensity, treatment, functional status), and sociodemographic factors (age, race, ethnicity, income, education, geographic location, access to care) predictive of the cLBP phenotype; and, (2) test the hypothesis that differential expression of MHC locus genes at baseline and over time will be associated with the risk of chronic pain, while differential expression of known pain genes will define the cLBP transcriptome.

2. METHODS

This longitudinal cohort protocol was prepared according to the SPIRIT 2013 guidelines (Chan et al., 2013) for completeness and quality of trial protocols and complies with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement (von Elm et al., 2014)

2.1. Design and setting

This prospective, longitudinal cohort study is recruiting subjects (n=340) with initial onset LBP from the University of Maryland, Baltimore (N = 170) and University of Connecticut (N = 170). Additionally, the study sample includes 40 age/race-matched healthy pain-free control participants (20 male, 20 female) who will be recruited from each site and undergo the same testing paradigm as the LBP group to normalize LBP participant data for analyses and track the stability of gene expression over time. The study has been approved by University of Maryland, Baltimore, Institutional Review Board.

2.2. Target population

Eligible participants are: (a) 18–50 years of age who comprehend English and report an acute episode of LBP present for >24 hours but ≤6 weeks duration and preceded by at least 6 months without an episode of LBP (de Vet et al., 2002). This age range was selected because it provides a more homogeneous sample for detecting transcriptomic predictors of cLBP as age >50 years is associated with a higher incidence of chronic conditions including arthritis, lumbar stenosis, and spondylolisthesis (Mehling et al., 2012). Detailed information related to analgesic use will be collected and incorporated into the data analyses. Potential subjects are excluded for the following conditions as provided by self-report during screening due to their potential to confound analysis of pain perception: (a) pain involving any other region of the body or associated with a chronic condition; (b) previous spinal surgery; (c) presence of neurological deficits or comorbidities that affect sensorimotor function; (d) recent fall within past three months; or, (e) history of unstable mental health conditions requiring medical management within the past 6 months. Healthy pain-free controls volunteers are excluded if they have pain in any region of the body, a history of low back pain, any surgical procedures or peripheral neuropathy, any history of a chronic health condition, or taking any type of medication regularly. Once enrolled, they will follow the same procedures as participants with low back pain.

2.3. Sample Size and Statistical Power

Previous cohort studies have found that at least 20% of patients 18 to 50 years old with new onset LBP reported chronic pain for 6 months or longer (Hestbaek, Leboeuf-Yde, & Manniche, 2003; Starkweather, Ramesh et al., 2016), thus we anticipate a similar proportion in the proposed study. Our targeted odds ratio (OR) of 1.6 represents a small to moderate effect size (Cohen, 1988). Several covariates were included in the sample size estimate, including somatosensory function, pain coping, mood, pain severity/interference, age, gender, race, ethnicity, and genetics. Correlations among covariates can be a major problem in multiple logistic regressions (Hosmer & Lemeshow, 2000; Menard, 2002). Therefore, Demidenko’s (2007) correction was employed in the power analysis by anticipating a medium-size intercorrelation among the covariates with the primary explanatory variables. Rosenthal (1996) and Valentine and Cooper (2003) discuss the limitations of using qualitative descriptors for effect sizes. Nevertheless, Cohen’s (1988) benchmarks are commonly employed because of their convenience and wide applicability.

The null hypothesis for the power analysis was that there would be no relationship between any one explanatory variable and the response variable (yes, no) of chronic pain. Assuming α = .05 with a two-tailed test, a conservative sample of N=266 will provide sufficient power (>.88) to detect a small effect size of association between chronic pain and an explanatory variable. Considering the potential attrition of about 25% (Mehling et al., 2012) our targeted enrollment will be 340 participants with acute LBP and 40 healthy (no-pain) controls for a total of N=380 participants (190 total at each site; 170 with LBP + 20 controls). Monte Carlo simulation showed that a sample of n=300 can identify up to 6 classes with 4 time points with sufficient power in latent class growth analysis (Kim., 2012). This study has a larger sample size and more data time points, thus the study will have sufficient power to conduct the the planned latent class growth analysis (Bi & Liu, 2016).

2.4. Procedures

University of Maryland Medical Center and University of Connecticut Health clinic staff identify potential participants during medical evaluation for initial onset LBP, present the study to the potential participant and ask if they are interested in talking with a member of the research team and if interested, they notify research staff. Research staff who are trained in informed consent procedures meet with the potential participant, describe the study and its objectives, and ask if they are interested in participating. Volunteers are screened using the inclusion and exclusion criteria described above. If eligible, subjects are given a clear explanation of what the study involves, risks and benefits to participation, the consent form to read over and a copy for their records. Participants are also given the option to consent to storage of their blood samples for use in further studies. The blood samples belonging to the participants who do not consent to this option will be discarded following the completion of this study. Research staff provide time to ask and answer questions and confirm that the potential participant understands their participation by asking questions such as, “Can you tell me your understanding of what you’d be asked to do if you participate in this study?” Research staff clarify or correct information about participation during this time After obtaining written consent, the participants are scheduled for their baseline clinic visit within one week. Healthy control participants are recruited via flyers placed at various locations in and around University of Maryland and University of Connecticut.

Data are collected by self-report questionnaires along with peripheral blood sampling for transcriptomic analysis and QST for somatosensory function during six in-person study visits over a two year time period (baseline, 6, 12, 24, 52 weeks, 2 years) (Table 1). During five remote time points (8, 10, 16, 20 weeks, and 18 months) only the self-report questionnaires are administered, which take approximately 30 minutes to complete. All questionnaires have been entered into REDCap^™ and are administered via iPad during the in-person testing visits and via phone or personal computer during the remote time points. The study time points were chosen based on critical points in the LBP trajectory and patterns of LBP recurrence over the 2-year time frame (de Vet et al., 2002; Mehling et al., 20112).

Table 1. Study Flow Diagram.

After consent is obtained, participants will complete the baseline questionnaires and undergo baseline QST as soon as possible within one week after initial onset visit. The questionnaires will then be completed during clinical testing visits at 6, 12, 24, 52 weeks, and 24 months. Participants will also complete the surveys on-line at home at 8, 10, 16, 20 weeks, and 18 months. Participants will undergo QST and blood draws during follow-up visits at 6, 12, 24, 52 weeks, and 24 months.

	BL	6 Wks	8 Wks	10 Wks	12 Wks	16 Wks	20 Wks	24 Wks	52 Wks	18 Mos	2 Yrs
Sociodemographic Data	X
Clinical Data	X	X			X			X	X		X
Brief pain Inventory	X	X	X	X	X	X	X	X	X	X	X
McGill Pain Questionnaire – Short Form	X	X	X	X	X	X	X	X	X	X	X
Coping Strategies Questionnaire	X	X	X	X	X	X	X	X	X	X	X
Kohn Reactivity Scale	X	X	X	X	X	X	X	X	X	X	X
Profile of Mood States	X	X	X	X	X	X	X	X	X	X	X
Roland Disability Questionnaire	X	X	X	X	X	X	X	X	X	X	X
Perceived Stress Scale	X	X	X	X	X	X	X	X	X	X	X
Neurophysiological Testing	X	X			X			X	X		X
Blood Draw	X	X			X			X	X		X

BL = Baseline; Wks = Weeks; Mos = Months; Yrs = Years; Orange = Clinical Testing Visit

2.5. Variables and measures

2.5.1. Clinical and sociodemographic characteristics.

Using a minimum data set recommended by the NIH Taskforce on Research Standards for Chronic Low Back Pain (Deyo et al., 2015), the following data is collected at the time of enrollment: (1) sociodemographic data (age, race, ethnicity, sex, gender, marital status, educational level, health insurance, primary care provider [yes/no], occupation, residence [urban/rural], smoking, alcohol, or use of controlled substances); (2) clinical data from self-report and medical record review (past and current medical history, medications, allergies, surgeries, date and time of LBP onset, past episodes of LBP, family history of LBP, height, weight, body mass index, pharmacological and non-pharmacological pain treatment). Items that may change over the course of the study (work history, substance use, treatments) are included at each study visit.

2.5.2. Self-reported pain characteristics.

Pain characteristics are collected with two measures, the Brief Pain Inventory – Short Form (BPI-SF) and Short-From McGill Pain Questionnaire-2 (SF-MPQ-2). The subscales within the BPI-SF assess the severity and location of pain, analgesic use and relief of pain during the past 24 hours and one week, as well as the impact of pain on activities of daily living. The BPI-SF is a reliable and valid instrument for assessment of clinical pain (Cleeland, 1991) and is sensitive to change (Keller, Bann, Dodd, Schein, Mendoza, & Cleeland, 2004). The SF-MPQ-2 assesses sensory and affective symptoms of acute and chronic pain, and overall pain intensity in adults (Clay et al., 2010; Dworkin et al., 2009). This self-report measure includes 15 pain descriptors that are scored on a 4-point scale of 0-none to 3-severe (Melzack, 1987; Melzack & Katz, 1991). The internal consistency (Cronbach’s alpha = 0.96), convergent validity and responsiveness to change of the SF-MP-2 have been reported as excellent (Lovejoy, Turk, & Morasco, 2012; Dworkin et al., 2015). In addition to the assessment of pain, we included the National Institute of Nursing Research (NINR) recommended common data elements for fatigue, sleep disturbance, anxiety and depression, as these symptoms commonly co-occur in this patient population and are of high interest to the field of symptom science.

2.5.3. Coping strategies.

The Coping Strategies Questionnaire - Revised (CSQ-R) is used to evaluate the participant’s coping strategies by self-report response to 27-items that assesses six cognitive coping responses to pain (Robinson et al., 1997). Participants are asked to rate the frequency that they use each coping strategy and their perceived control over the pain using a 7-point Likert-type scale that ranges from “never do that” to “always do that”. The CSQ-R subscales have demonstrated adequate internal consistency (Cronbach’s alpha from .72 to .91) (Hastie, Riley, & Fillingim, 2004) and stable factor structure (Riley & Robinson, 1997).

2.5.4. Reactivity.

The Kohn Reactivity Scale (KRS) (Kohn, 1985) is used to assess the participant’s central nervous system arousability (McDermid, Rollman, & McCain, 1996). Participants self-report their reactivity in this 24-item measure. After reverse scoring half of the items, the KRS yields a sum score of all the individual items. The KRS has demonstrated adequate internal consistency with a Cronbach’s alpha ranging from .73 to .83 (Kohn, 1985) and a negative correlation with pain tolerance (Dubreuil & Kohn, 1986).

2.5.5. Mood.

The Profile of Mood States (POMS) measures the participant’s general distress and mood (McNair & Lorr, 1964; McNair, Lorr, & Droppleman, 1971). The POMS consists of 65 items with 6 subscales: tension-anxiety, depression-dejection, anger-hostility, fatigue, vigor, and confusion-bewilderment. Participants respond to each item using a 5-point Likert-type scale ranging from 1 (not at all) to 5 (extremely). After summing the scores of each subscale a total mood disturbance score is obtained (range −40 to 192). The POMS has demonstrated good internal consistency with Cronbach’s alpha ranging from .87 to .95 and test-retest coefficients of .65–.74 (McNair & Lorr, 1964; McNair, Lorr, & Droppleman, 1971).

2.5.6. Disability.

The Roland Disability Questionnaire (RDQ) is used to assess perceived LBP-related disability. The RDQ is the most widely used tool to measure perceived disability in the LBP population (Roland & Morris, 1983) and is sensitive to change (Kopec & Esdaile, 1995). The RDQ score is derived by summing checked items. The test-retest reliability, content and construct validity, and internal and external responsiveness have all been shown to be satisfactory (Davies & Nitz, 2013).

2.5.7. Perceived Stress.

The Perceived Stress Scale (PSS) measures the participant’s level of stress and consists of 10 items (Cohen, 1994). The PSS asks participants to appraise the stressfulness of situations in their lives over the past month on a five-point scale from ‘never’ to ‘very often’. To calculate a total PSS score, four positively stated items are reverse scored and then all items are summed. The instrument has been demonstrated to be reliable and valid (Kain, Sevarino, Alexander, Pincus, & Mayes, 2000).

2.5.8. Quantitative Sensory Testing.

During clinical testing visits, QST is used to examine nociceptive and non-nociceptive thermal and mechanical sensory function. Figure 1 provides an overview of the testing paradigm and test order implemented that is based on the standardized procedures of the German Research Network (Rolke, Baron et al., 2006; Rolke, Magerl et al., 2006). Thermal testing is delivered by a Pathway thermal stimulator (Medoc; Ramat, Israel) equipped with two thermodes (ATS [2.56cm²], CHEPS [5.73cm²]) and safety cutoff temperatures of 0°C and 50°C. Testing is performed at the inner aspect of the non-dominant forearm as the control site followed by the most painful area of the lumbar region. Participants rate their pain using the verbal numeric rating scale (NRS) with “0” indicating “no pain” and “100” indicating “worst pain imaginable”.

Figure 1. Order of Neurophysiological Testing.

CDT=cool detection threshold; WDT=warmth detection threshold; TSLT=thermal sensory limen test; CPT=cold pain threshold; HPT=heat pain threshold; MDT=mechanical detection threshold; DMA=dynamic mechanical allodynia; PPD=pin prick detection; VDT=vibration detection threshold; MTS=mechanical temporal summation; PPT=pressure pain threshold; CPM=conditioned pain modulation.

a. Thermal Detection and Thermal Pain Thresholds (TD & TPT): is done to determine thermal sensory loss (hypoalgesia) or sensory gain (hyperalgesia) and uses a “method of limits” protocol common to many pain-testing laboratories, including the German Research Network (Rolke, Baron et al., 2006; Rolke, Magerl et al., 2006). To estimate each of the four thresholds, the participant is asked to press a button with the dominant hand as soon as they experience the sensation of interest for a given test (cool sensation (cool detection threshold [CDT]), warm sensation (warm detection threshold [WDT]), cold pain (cold pain threshold [CPT]), heat pain (heat pain threshold [HPT]). The test consists of gradual heating or cooling of the ATS thermode (resting temp. 32°C, ramp 1°C/sec).

b. Thermal Sensory Limen Testing (TSLT): A diverse array of neurological disorders exhibit paradoxical heat sensation to cool stimuli (Susser, Sprecher, & Yarnitsky, 1999). To examine whether participants exhibit this phenomenon, six alternating hot or cold stimuli are delivered (resting temp. 32°C, ramp 20°C/sec). After each stimulus, the participant responses are recorded (Rolke, Baron et al., 2006; Rolke, Magerl et al., 2006).

c. Mechanical Detection Threshold (MDT): is performed to determine mechanical sensory loss (hypoalgesia) or sensory gain (hyperalgesia). It is conducted using a set of 20 Semmes Weinstein monofilaments (Touch Test Sensory Kit, myNeurolab.com) with varying diameter, equal length, and calibrated to bend at a specified amount of force (0.008 gm – 300 gm). Testing begins at the 0.07 gm fiber and continues to the next larger fiber until the fiber is detected in 2 out of 3 applications, which is recorded as the MDT.

d. Dynamic Mechanical Allodynia (DMA): is done to assess mechanical sensory gain resulting in the development of allodynia (a painful response to a normally non-painful stimulus). This test is done using a standardized light tactile brush stimulator (Somedic; Sweden) (Baumgärtner, Magerl, Klein, Hopf, & Treede, 2002). The stimulus is applied three times and the mean of the three NRS pain scores is then calculated.

e. Pin Prick Detection (PPD): is performed to assess mechanical sensory function (Rolke, Baron et al., 2006; Rolke, Magerl et al., 2006). It is conducted with the participant’s eyes closed using an aseptic pin (sharp) and an aseptic paper clip with one end bent 90° from the clip body to form a probe (dull). Three sharp and three dull stimuli are randomly applied causing a slight indentation of the skin but no puncture. The participant reports whether the sensation is sharp or dull. The stimulus type and response are recorded.

f. Vibration Detection Threshold (VDT): is used to detect abnormality in sensory quality in peripheral nerves (Vagter, Palsson, & Graven-Nielsen, 2017) and is determined using a graduated tuning fork (Rydel-Seiffer, US Neurologicals) fitted with calibrated weights at the ends that vibrate at 64 Hz (Martina et al., 1998). The participant reports when the vibration is no longer felt and the investigator records the number on the calibrated weight nearest to the intersection of the triangles. A mean score is calculated from the three trials.

g. Mechanical Temporal Summation (MTS): Wind-up is a phenomenon involving increased spinal cord neuron excitability in response to repetitive mechanical stimuli (Herrero., Laird Lopez-Garcia, 2000) and can result in an increase of pain (Kong, Johnson, Balise & Mackey, 2013). The pain rating of a single application of a sharp 256mN weighted probe is determined, followed by a series of ten repetitive applications (1/sec) of the same probe. The participant then rates the overall pain from the series of ten stimuli. This paradigm is repeated for five trials. The wind-up ratio is the perceptual correlate of MTS and is calculated by dividing the mean rating of the five series trials by the mean rating of the five single stimulus trials (Rolke, Baron et al., 2006; Rolke, Magerl et al., 2006).

h. Pressure Pain Threshold (PPT): quantifies soft tissue tenderness and identifies the presence of hyperalgesia. It is determined using a pressure algometer (Medoc; Ramat, Israel) with a 0.5cm blunt rubber tip. The pressure algometer is manually applied at a steadily increasing force of 30 kPa/sec until the participant presses a button to indicate onset of pain. The mean of the two closest values in the five trials is calculated to estimate the PPT (Greenspan et al., 2011), with a maximum cutoff pressure of 600 kPa for safety.

i. Conditioned Pain Modulation: measures the function of the endogenous descending pain modulatory circuit and is assessed by delivering a test stimulus (noxious heat at a fixed pain intensity of 40 on a 0–100 numeric pain rating scale) to the lower back and a conditioning stimulus to the non-dominant hand (Kosek & Hansson, 1997; Yarnisky, 2010; Yarnitsky, Arendt-Nielsen et al., 2010). The CPM score is the change in the numeric pain rating scale score from the test stimulus that is induced by the conditioned stimulus. It is reported using both changes in the absolute values and percent change of the sensation or physical values.

2.5.9. Blood collection for RNA sequencing.

In order to assess transcriptomic changes over time, we collect peripheral blood from all participants using Tempus Blood RNA collection tubes, which allow for prolonged stability during clinical testing visits. Per the manufacturer’s instructions, research staff at University of Maryland, Baltimore, will isolate and purify total RNA from the whole blood. The blood samples are stored at −80° Celsius for bulk processing.

2.6. Planned analysis

We will apply a series analysis approach to identify the neurophysiological predictors of the transition from acute to chronic pain. We will use: (1) multiple logistic regression analysis to estimate the odds of developing chronic pain on each of the explanatory variables [i.e., physiological, psychological, clinical factors] at baseline, controlling for other predictors/covariates in the model; (2) latent class growth analysis (LCGA) (Nagin & Land, 1993; Muthen, 2004) to examine the relationship between cLBP and the trajectories (latent intercept and slopes) of the explanatory variables from baseline to 6 months. LCGA is recognized for its usefulness of identifying homogeneous subpopulations within the larger heterogeneous population and for capturing inter-individual differences in intra-individual changes over time (Jung, 2008). The trajectories of pain intensity or neurophysiological scores will be modelled by latent intercept (e.g., baseline mean score of pain) and latent slopes (e.g., changes of pain scores). Each participant will then be classified into different homogeneous classes based on the trajectories (i.e., latent intercept and slopes), along with time invariant covariates (e.g., demographics) and time varying covariates [e.g., physiological and psychological factors]. The associations between cLBP and the trajectories within each homogeneous class will be estimated.

The Lo, Mendell and Rubin (LMR) likelihood ratio test (LRT) and Bootstrap LRT will be used to determine the number of classes in mixture modeling, along with the consideration of other model fit criteria such as Akaike information criterion (AIC), Bayesian information criterion (BIC), entropy as well as theoretical justification and interpretability (Nylund, Asparouhov & Muthen, 2007). While we do not plan to adjust for multiple hypothesis testing, we will follow best practices in analytic reporting. Once the class membership is identified, we will further assess the associations between class membership and the covariates and development of chronic pain, which include the patients’ demographics, clinic data, and other questionnaire measures. The LCGAs will be performed using Mplus 8 (Muthen & Muthen, 2017). Parameter estimation in LCGAs is carried out with robust maximum-likelihood and the expectation-maximization (EM) algorithm. Missing data in all models are managed with the full information maximum likelihood (FIML) procedure used by Mplus (Muthen & Muthen, 2017). Finally, we will extend growth curve models in each latent class to time points from six months to two years follow-up period to examine long term development of pain, neurophysiological and other measures, especially among those with persistent pain for up to two years.

2.6.1. Transcriptomic analysis

In order to identify differently expressed genes associated with neurophysiological and acute/chronic pain phenotypes the same workflow used in prior published studies (Dorsey et al., 2019). Briefly, libraries will be prepared per Illumina standard protocols and the Illumina HiSeq2500 System (Illumina Inc.) will be used to sequence RNA to obtain 150–base pair paired-end reads. Each sample will be barcoded and we will multiplex them on a lane (n=4–6) to obtain 40–80M reads per sample, the gold standard for differential gene expression studies (Mortazavi, Williams, McCue, Schaeffer, & Wold, 2008). Reads will be mapped to the most recent build of the Ensembl human genome with the short-read aligner TopHat (Trapnell, Pachter, & Salzberg, 2009) and alignments will be merged to form complete alignments to the genome. Using the count method, the number of reads mapped on each gene will be counted with the HTSeq program (Anders, Pyl, & Huber, 2015) against the gene annotation file for the Ensembl human genome build (Yates et al., 2016) to determine differential gene expression by RNA-seq. The R package DESeq (Anders & Huber, 2010) will be used for data normalization and differential expression analysis. To test the significance of differential expression between two or more conditions a negative binomial model is used at a False Discovery Rate (FDR) cutoff of less than 0.05 with a more than twofold change in gene expression between conditions and a normalized read count of 10 reads per gene in at least one condition.

3. Discussion

In this study we will build upon our previous research and follow participants with a new onset of LBP as they either recover or transition to cLBP. The data obtained from participants will be used to fill a wide gap in knowledge concerning the risk factors and recurrence of axial LBP over time. The large sample size (LBP patients: n= 340; Controls: n=40) and use of age matched controls will allow us to properly carry out analyses aimed at identifying genotypic and phenotypic differences between individuals. To our knowledge, there is only one similar study, however, data is only collected only at 3, 6, 12, and 24 months and only a subsection of participants receive neurophysiological and genetic testing (Page et al., 2020). In addition, our study will be the first to advance understanding about the neurophysiological and genomic changes that occur beyond the initial stages of pain chronicity (6–12 months). Lastly, this study will follow a protocol similar to our ongoing study which aims to identify neurophysiological, and genomic factors of chronic pain following a lower extremity fracture (Griffioen et al., 2020). Aside from minor changes between QST measures to account for differences in clinical presentation, both protocols outline nearly identical exclusion criteria, study visit intervals, and data collection methods. These similarities will reduce potential confounders and allow us to compare predictors of chronic pain across pain conditions.

As the COVID-19 pandemic evolved around the same time as the initial launch of study recruitment, our team was challenged to identify different recruitment and data collection procedures, including use of social media for recruitment and screening, routine COVID-19 testing for research personnel, and implementation of social distancing while performing data collection. A COVID-19 screening protocol was put into place prior to in-person data collection visits along with questionnaire items on COVID-19 infection and vaccination history.

At the start of the study, we planned to recruit participants through clinical referral and fliers placed at and near urgent and primary care clinics. Since the COVID-19 pandemic restricted access to clinical sites and decreased traffic at nearby locations, advertisements were placed on affiliated websites, social media, and sent through email lists to reach potential participants. Electronic advertisements have the advantage of easily directing potential participants to complete a secured electronic screening form. We have learned that compared to clinical referrals and on-site advertisements, which require potential participants to contact research personnel for initial screening, the electronic screening method significantly reduces recruitment burden on study staff.

Despite these measures, recruitment continues to be a challenge. To date, 665 people have been screened with 425 ineligible due to ongoing cLBP. Of those eligible, 224 did not enroll due to COVID-19 (+) status, difficulty with scheduling or transportation, and/or lost interest in participating. In total, 16 subjects were consented and completed baseline data collection. The recruitment methods and results are reviewed each week and our team will be focused on making modifications as needed to recruit a diverse sample of participants, with the hope that increasing COVID vaccination rates will ease hesitancy in clinical research participation.

4.0. Limitations

Our research team has a breadth of experience in conducting longitudinal studies using biopsychosocial, RNA sequencing, and QST. However, as this is a multi-site study, we are aware that interrater differences can be of some concern, particularly when conducting QST. To reduce these biases, we conducted an in-person training session so that researchers performing QST are strictly following the same protocol. To minimize potential batch effects, all RNA extraction and sequencing will be conducted at the University of Maryland Baltimore. Furthermore, use of healthy controls will be used to track the occurrence of batch effects and any differences that can be attributed to the use of new technologies during the study. In regard to translatability, we acknowledge that the use of the full QST protocol in the clinical setting is unrealistic due to the time required. However, the goal of this project is to identify the tests that are highly predictive of cLBP so that only a limited number will be needed to provide a risk assessment in the clinical setting to promote early intervention. In addition, the accessibility and cost associated with RNA sequencing would not be feasible at this time, but our goal for this program of research is to advance steps toward future point-of-care testing to enable more expedient diagnosis and treatment of axial LBP.

5.0. Conclusions

This ongoing prospective, descriptive cohort study aims to identify a comprehensive transcriptomic signature of cLBP and identify neurological and transcriptomic biomarkers which can predict cLBP at LBP onset. Along with documenting the neurophysiologic changes that occur during the transition from acute to cLBP, transcriptomic alterations may provide a powerful tool for identifying the molecular events that underlie pain resolution and chronicity.