Baseline characteristics of participants in the Biomarkers for Evaluating Spine Treatments clinical trial: a sequential multiple assignment randomized trial for chronic low back pain

Abstract

Objective

Chronic low back pain (cLBP) is a significant public health problem in the United States. A method to identify treatments that are most likely effective for an individual patient based on their unique characteristics is needed.

Methods

The Biomarkers for Evaluating Spine Treatments (BEST) Trial is a sequential, multiple assignment, randomized trial designed to estimate an optimal treatment or combination of treatments to reduce pain intensity and interference at 24 weeks in individuals with cLBP.

Results

We describe the patient-reported characteristics of the BEST Trial at the Baseline visit. Data collection for extensive required phenotyping is reported. We analyzed the run-in period of the BEST Trial to evaluate predictors of run-in failure. The BEST Trial enrolled 1019 participants and randomized 805 participants (61.6% female, mean age 50.4, 12.5% Black or African American) to the first stage of treatment. We collected extensive required phenotyping on all 805 randomized BEST Trial participants, and additional optional phenotyping on 510 (63.4%) participants.

Conclusions

The BEST Trial successfully enrolled a racially and geographically diverse sample of chronic low back pain patients and completed rich phenotypic assessments to inform our primary goal of identifying in whom different treatments show optimal response. We demonstrated the feasibility of collecting extensive phenotypic assessments in a multi-site clinical trial of cLBP.

Clinical trial registration number

The Biomarkers for Evaluating Spine Treatments (BEST) Trial is registered on ClinicalTrials.gov. Registration number: NCT05396014 (https://clinicaltrials.gov/study/NCT05396014).

Introduction

Chronic low back pain (cLBP) is a significant public health problem that impacts quality of life and reduces physical function.¹ cLBP is the leading cause of disability in the United States.² In addition to a tremendous impact on human suffering, cLBP results in an estimated 96-500 million dollars in United States healthcare expenditures annually.^3,4 Previous studies have established effective treatment strategies for cLBP.^5,6 However, individual patients are frequently prescribed a treatment that is ineffective for them and often try multiple treatments. During this treatment process, some patients may find pain relief while many continue to report disabling lower back pain despite numerous treatments.^7–9 To improve long-term outcomes in cLBP patients, a method for identifying treatments that are most effective for each patient is urgently needed.

One potential approach to improve the treatment of cLBP is to identify biomarker-based characteristics of cLBP patients and tailor treatment accordingly, a precision medicine approach. Biomarkers of cLBP have been identified across a wide range of measurement domains: psychosocial, biomechanical, ‘omics, inflammatory, spine and brain imaging, and social determinants of health. However, these biomarkers have neither been evaluated comprehensively nor demonstrated the ability to tailor treatment assignments to participants in clinical studies.

In response to the need to improve treatment outcomes for cLBP patients, the National Institutes of Health (NIH) Helping to End Addiction Long-term (HEAL) Initiative and the Back Pain Consortium (BACPAC)¹⁰ designed the Biomarkers for Evaluating Spine Treatments (BEST) Trial (NCT05396014). The BEST Trial was a sequential, multiple assignment, randomized trial (SMART)^11,12 designed to estimate the optimal treatment or combination of treatments for an individual participant based on phenotypic biomarkers and an individual’s response to the initial treatment13 (Figure S1). The motivating hypothesis of the BEST clinical trial is that there exist biomarkers that will define clinically relevant patient phenotypes and allow individual tailoring to improve cLBP treatment. Data collection for the BEST Trial was designed to answer this hypothesis by collecting a comprehensive set of state-of-the-art biomarkers across measurement domains to maximize the likelihood of discovering tailoring biomarkers (Table 1).^14–25 The protocol and methods used have been described elsewhere.^13,26–32 The interventions included in the BEST Trial were (1) Enhanced Self Care (ESC), (2) Acceptance and Commitment Therapy (ACT), (3) Evidence-Based Exercise and Manual Therapy (EBEM), and (4) Duloxetine,¹⁰ which have all been previously shown to be effective for the treatment of cLBP. The statistical design of the trial was motivated by the aim of biomarker discovery without the need for pre-specifying hypotheses about specific biomarkers using modern precision medicine approaches. It will be described in detail in a forthcoming statistical design manuscript.³²

Table 1.

Rationale for including biomarker domains in BEST Trial data collection.

Biomarker domain	Justification for inclusion	Citation
Psychosocial	Anxiety and depression have been associated with pain interference in cLBP patients.	14
Movement and Muscular Biomechanics	Movement and muscular biomechanics biomarkers can differentiate cLBP patients from healthy controls.	15
Molecular	Single nucleotide polymorphisms (SNPs), gene expression profiles, and DNA methylation each have implicated known pain genes and inflammatory signaling.	16–18
Inflammatory	Pro-inflammatory biomarkers demonstrate a positive correlation with low back pain such as C-reactive protein, interleukin 1 and IL-1β, interleukin 6 (IL-6), and tumor necrosis factor alpha (TNF-α).	19 , 20
Blood-based markers	Whole blood was collected to perform immune profiling via mass cytometry and to perform TruCulture assays enabling immune stimulation given relationship between pain and immune cell activation.	21
Stool	Analysis of gut microbiome provides fundamental insights on connections between the microbe composition and chronic pain characteristics such as systemic inflammation, anxiety, and depression.	22
Spine Imaging	Low back pain patients with high intensity zones (HIZs) on lumbar magnetic resonance imaging (MRI) findings were more likely to have disc degeneration/displacement and were associated with severe LBP and sciatica.	23
Brain Imaging	Functional MRI (fMRI) technology has elucidated the biomarker potential of large-scale functional reorganization in chronic pain patients.	24
Social Determinants of Health	Social determinants of health such as educational attainment and socioeconomic status are associated with cLBP.	25

This manuscript describes the recruitment process and baseline characteristics of the participants enrolled in the BEST Trial. Furthermore, we conduct an exploratory analysis to identify factors that predicted non-completion during the run-in period to evaluate this component of the study design and the generalizability of our findings to inform future trials. For future study planning, we present combinations of participants with completed required and optional phenotypic assessments. Because of our comprehensive biomarker collection, we anticipate the BEST Trial being a rich dataset for future analyses. The baseline results of the BEST Trial highlight the epidemiological characteristics of the trial cohort in which a comprehensive set of biomarker assessments have been completed.

Methods

BEST Trial design

Study aims

The primary objective of the BEST Trial was to inform a precision medicine approach to treating cLBP by estimating the optimal treatment or combination of treatments for an individual participant based on patient features and an individual’s response to the initial treatment.^13,32

Recruitment

Potentially eligible participants were identified by direct contact in clinical settings, screening via electronic health records for cLBP, online surveys, and community outreach, including social media. Inclusion and exclusion criteria were designed to avoid being overly restrictive in enrolling a representative population and to allow for generalizability to more individuals with cLBP. Detailed descriptions of the inclusion and exclusion criteria have been described elsewhere,¹³ and summaries of eligibility criteria related to recruitment are included in Table S1. Potential participants meeting eligibility criteria were pre-screened into the trial via phone and scheduled for an enrollment visit with screening assessments approximately 3 weeks before visit 0 (Baseline). All enrolled participants entered a 2-week run-in period. The goal of the run-in period was to engage enrolled participants, evaluate their adherence, and enhance retention.¹³ Early in the recruitment period of the study, the Executive Leadership Committee approved a protocol modification effective November 15, 2022 to drop the number of daily pain questionnaires that were needed to be completed by participants during the run in period from 10 to 5 since participants were excluded from the study because of technical issues while being highly engaged. For example, the daily pain questionnaires went to their junk mail without the participants' knowledge or failed to deliver the surveys because of computing errors. Additionally, this modification gave participants more time to view educational modules required for randomization. These modules could be completed with the coordinator at the randomization visit rather than needing to be completed before the visit. On July 13, 2023, the threshold of daily pain questionnaires dropped to 0. Despite further modifications to increase the ease of maintaining eligibility in the run-in period, 154 participants who had consented to the trial were still lost during the study period due to withdrawal or loss to follow-up. After the run-in period, participant eligibility was re-evaluated, and those meeting criteria were randomized and began the stage 1 intervention.

Study outcomes

The primary endpoint of the BEST Trial is the 24-week Pain, Enjoyment of Life and General Activity (PEG) Scale score, a mean composite score of the responses to three 0-10 patient-reported outcomes (PROs) regarding 3 domains of pain experience: pain intensity, pain interference with activity, and pain’s impact on the enjoyment of daily life. The secondary endpoints are 24-week pain interference, the incidence of opioid use measured at 24 weeks, 24-week physical function, 24-week depression score, 24-week anxiety score, 24-week sleep disturbance, and 24-week sleep duration.¹³

Trial operating characteristics

The BEST Trial used a sequential multiple assignment randomized trial (SMART) design at 11 study sites (Figures S1 and S2). The trial was divided into two 12-week intervention stages with 3 study visits that occurred at the beginning of stage 1 (visit 0/week 0), stage 2 (visit 1/week 12), and the end of stage 2 (visit 2/week 24). Participants were randomized to 1 of the 4 study interventions at baseline (visit 0/week 0). Participants must have been eligible for at least 3 out of the 4 study treatments at baseline. The week 12 re-randomization procedure has been described extensively elsewhere.¹³ We briefly summarize here. At week 12 (visit 1), participants were designated to 1 of 4 responder status categories using a combination of week 12 Pain, Enjoyment of Life, and General Activity (PEG)³³ and Patient Global Impression of Change (PGIC)^34,35 scores (Figure S3). Participants with week 12 PGIC 1-2 and PEG ≤ 4 maintained their initial treatment, participants with week 12 PGIC 1-2 and PEG > 4 were randomized to augment a second treatment, participants with a week 12 PGIC 3-4 were randomized to either augment a second treatment or switch treatments, and participants with a week 12 PGIC of 5-7 or who could not tolerate their stage 1 treatment were switched to a new treatment.¹³ Randomization was stratified by years of chronic pain, consent to optional phenotyping, anxiety/depression, and current self-reported opioid use using a modified randomization and minimization algorithm, which maintains balance despite study participants not being eligible for all treatments at randomization.

Interventions

All BEST Trial interventions have been established as effective treatment strategies for cLBP and are feasible to execute in a multi-center study.⁶ Patients, assessors, and providers were unblinded to treatment assignment since the trial's primary aim was biomarker discovery. ACT included exposure to 12 weeks of acceptance and commitment therapy through 4 video visits with a therapist and 8 therapist-supported asynchronous online sessions. EBEM included 10 in-person sessions that included a combination of exercise, manual therapy, assessment of psychosocial factors, and participant education for spinal care. Duloxetine is a serotonin-norepinephrine reuptake inhibitor (SNRI) approved by the Food and Drug Administration (FDA) for use in Chronic Musculoskeletal Pain such as cLBP. Duloxetine was prescribed at an initial dose of 30 mg and increased to 60 mg after 1 week, conditional on tolerance of the medication. ESC was an asynchronous online module that included information on cognitive-behavioral self-management skills for pain and an embedded walking program using Fitbit step tracking.

Study assessments

All participants participated in required phenotyping at Baseline (week 0), week 12, and week 24, which consisted of answering questionnaires, receiving physical and short motion assessments, and providing blood samples at all 3 time points.¹³ At baseline (week 0), participants provided one stool sample and received one spine MRI. PRO assessments were performed at Baseline and weeks 6, 12, 18, 24, and 36. Outcomes for the primary aim were collected at week 24.

A subset of eligible and willing participants completed additional optional phenotyping at Baseline (week 0), week 12, and week 24. These optional assessments included more comprehensive motion assessments, advanced brain MRI, advanced spine MRI, and Quantitative Sensory Testing at each visit. Furthermore, at only the baseline visit, these participants were asked to continuously wear an actigraphy sensor at home for 7 days to assess their physical activity.

Ethics and external trial review

Participant safety was overseen by a Data and Safety Monitoring Board (DSMB) chartered by the National Institutes of Health (NIH) and the National Institute of Arthritis and Musculoskeletal and Skin Diseases (NIAMS) and administered by Navitas Clinical Research, Incorporated. The BEST Trial utilized Advarra, Inc. as the trial's single Institutional Review Board (IRB).

Statistical analysis for the baseline manuscript

We performed a descriptive analysis of the baseline characteristics of randomized BEST Trial participants. We summarized quantitative data using mean and standard deviation and qualitative data using counts and percentages. The PROs utilized in this analysis have been described elsewhere.¹³ Upset plots were used to describe overlapping relationships between sets of participants, particularly for treatment eligibility and data collection.³⁶

Participants who provided consent but failed the initial run-in period were studied to develop a predictive model for run-in failure. A binary outcome was used to analyze all reasons for run-in failure, including exclusion, loss to follow-up, and study withdrawal. The study design change for the requirements for the daily pain questionnaires was accounted for with a categorical variable with 3 levels based on the date entering their run-in period and the current protocol requirements (see Methods—Recruitment section). Any missing data were imputed using random forest imputation with 1000 trees and 50 iterations, assuming the data are missing at random (Figure S4³⁷). Subsequently, a random forest model was applied to all survey data collected during the run-in period, encompassing demographic information, pain assessments at pre-screening, social determinants of health, and PROs.^38,39 The final random forest model excluded categorical variables that were constants after imputation. The dataset was then randomly split into 80% training and 20% testing sets. The random forest model was trained on the training set and evaluated on the testing set. Variable importance from the trained model was calculated using mean decrease Gini. Mean decrease Gini is a measure of variable importance where higher values indicate that the variable is more useful for predicting run-in failure for the random forest model. For the top 10 most important variables based on mean decrease Gini, partial dependence plots were used to examine the univariate relationship between the variable and the probability of run-in success.^40,41

Results

Participant screening and enrollment

We pre-screened 3140 potential participants, screened 1483 potential participants, enrolled 1019 participants, and randomized 805 participants in the BEST Trial (Figure 1). Participants were recruited from a nationwide, geographically diverse set of trial sites (Figure S2). We observed ratios of 3.9:1 and 1.8:1 for pre-screened and screened potential participants to those who were eventually randomized. Most cLBP participants were identified by recruitment from online/print advertising or directly recruited from clinical settings.

Figure 1.

CONSORT diagram through stage 1 randomization.

The most common reasons for pre-screen ineligibility were low PEG score (584/1408), a disqualifying medical condition (429/1408), back pain was not more severe than pain in other parts of the body (334/1408), or back pain for less than 3 months (204/1408) (Figure 1). The most common reason for screen failure was not being eligible for 3 out of the 4 study treatments (103/280). Seventy-two participants became ineligible during the run-in period, and 82 were lost to follow-up. The most common reasons for exclusion after the run-in period were lost to follow-up participants, failure to complete daily pain questionnaires, and failure to complete the BEST Trial educational modules.

Using a random forest model for exploratory analysis, we identified variables associated with eligibility after the run-in period (see Methods section). Using a 20% testing set of 203 participants, we estimated the random forest model’s accuracy, sensitivity, and specificity to be 91.6%, 98.9%, and 46.4%, respectively. Variable importance analysis revealed that self-reported income, PROMIS Cognitive Function 2a raw score, and the PROMIS Depression 4a T-score were most important for predicting run-in failure (Figure S5A¹³). Moderate income levels, higher levels of cognitive function, and absence of depression symptoms were associated with a higher probability of eligibility after run-in (Figure S5B).

Treatment eligibility among those who were randomized to the study showed that 346 (43.0%) of participants were eligible for all 4 study treatments, 345 (42.9%) were ineligible for duloxetine, 112 (13.9%) were ineligible for EBEM, and 2 (0.2%) were ineligible for ACT (Figure 2). The most common reasons for ineligibility to duloxetine were concomitant medications that could potentially interact adversely with duloxetine (n = 257; 31.9%), participant history of bipolar disorder, manic episodes, or suicide attempts (n = 59; 7.3%), and history of an allergic or serious adverse reaction to duloxetine (n = 59; 7.3%). The most common medication classes that rendered participants ineligible for duloxetine were selective serotonin reuptake inhibitors (n = 154) and serotonin-norepinephrine reuptake inhibitors indicated for pain or depression (n = 95), including duloxetine (n = 46) (T4) (Table S2A). The most common reasons for EBEM ineligibility were participants receiving or intended to receive any manual therapy or exercise for their cLBP (n = 46), as well as shortness of breath or chest pain (n = 27), and a failed clinical assessment (a sustained systolic blood pressure (SBP) ≥ 150 or diastolic blood pressure ≥ 100) (n = 26) (Table S2B). The reason for ineligibility to ACT was participants receiving or intending to receive pain-specific psychotherapy (n = 2) (Table S2C). All participants were eligible at baseline for ESC.

Figure 2.

Treatment eligibility at baseline. The BEST Trial was designed such that a participant had to be eligible for only 3 out of the 4 treatments to be randomized at the baseline visit. During both randomization procedures at week 0 and week 12, participants could only be randomized to treatments for which they were eligible. The upset plot (Figure 2) illustrates the combinations of treatment eligibility among participants randomized at the baseline visit. The vertical bars count the number of participants who had the same set of treatments for which they were eligible. The sets of treatments are defined by the solid dots connected by lines below the vertical bar. For example, 346 randomized participants (43.0%) were eligible for all study treatments at the baseline visit. 345 participants (42.9%) were eligible for EBEM, ACT, and ESC and ineligible for Duloxetine. The horizontal bars count the number of randomized participants eligible for each treatment individually. 460 randomized participants (57.1%) were eligible for Duloxetine, 694 participants (86.2%) were eligible for EBEM, 803 (99.8%) participants were eligible for ACT, and 805 participants (100%) were eligible for ESC. In the primary analysis of the BEST Trial, we will account for the set of possible treatments that someone could have received when identifying biomarkers that predict treatment response by defining the correct subject-specific eligible set of treatments.

Baseline characteristics

At baseline, the 805 participants randomized to the first treatment stage had a mean (SD) age of 50.4 (SD = 16.1) years, 59.6% were women (n = 495), 71.1% were White (n = 572), 12.5% were Black or African American or Black multiracial (n = 101), and 8.9% were Hispanic or Latino (n = 72) (Table 2). Most participants had a college degree (63.9%), and 35.7% had a household income greater than $100 000. The average duration of chronic pain among randomized participants was 13.2 (SD = 11.8) years, the average baseline PEG score was 5.3 (SD = 1.9), and the average low back-specific pain intensity was 5.6 (SD = 1.8). 83.6% of participants reported having lower back pain every day or nearly every day in the past 6 months. Self-identified regions of pain on the Michigan Body Map demonstrated that our cohort has chronic pain in the area of the lower back from the posterior margin of the rib cage to the iliac crest (96.9%), as well as buttocks (L: 51.6%, R: 53.8%), neck pain (44.1%), and knee (L: 25.5%, R : 26.7%) (Figure 3A). Participants often experienced pain in multiple distinct pain zones (Figure 3B; zones are defined as left arm, right arm, left leg, right leg, chest, back, and head with neck). Among participants, 12.9% had at least 1 lower back surgery, 18.1% had experienced unemployment for 1 month because of lower back pain, and 10.4% had received or applied for disability insurance for a pain condition. Many participants had self-reports of conditions associated with chronic pain via PROMIS measures, including higher than average levels of pain interference [61.7, (5.5)], fatigue [56.8, (9.2)], sleep disturbance [55.4, (7.8)], anxiety [53.2, (9.4)], and depression [52.7, (9.4)], as well as lower than average levels of social participation [45.1, (7.1)] and physical function [40.2, (5.7)] (Figure 3C). Participants self-reported that physical function, pain interference, and pain intensity were the most important outcomes to improve regarding limitations of their back pain and that current levels were unsatisfactory (Figure 3C and D). Nearly 70% of participants had moderate or severe disability as measured by Oswestry Disability Index, and 63% had nociceptive pain, with a lesser proportion, 13%, experiencing neuropathic pain as measured by the PainDetect Questionnaire. Over the 3 months before the baseline visit, participants most commonly were engaged in an exercise routine (n = 419; 52%), used NSAIDS (n = 290; 36%), tried a diet/nutrition program (n = 199; 24.7%), engaged in mindfulness or meditation (n = 158; 19.6%) to treat chronic low back pain (Figure S6, Table S3). In the previous 3 months, 51 participants (6.3%) had used opioids at all, and 5.1% were currently using opioids at the baseline visit. The most common combination of treatments participants received in the 3 months leading up to the baseline visit were NSAIDs and Exercise in combination (n = 54, 6.7%). Physical function was assessed on all randomized participants at baseline (Table S4). We did not observe deviations from balance in the randomization procedure.

Table 2.

Tabular characteristics of randomized population.

Variablea	ACT (N = 203)	Duloxetine (N = 198)	EBEM (N = 199)	ESC (N = 205)	Total (N = 805)
Age	49.4 (16.5)	51.6 (15.4)	49.5 (16.3)	51.2 (16.1)	50.4 (16.1)
Sex assigned at birth, n (%) b
Female	116 (57.1%)	115 (58.1%)	131 (65.8%)	134 (65.4%)	496 (61.6%)
Male	87 (42.9%)	83 (41.9%)	68 (34.2%)	71 (34.6%)	309 (38.4%)
Missing	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)
Gender identity, n (%)
Female	112 (55.2%)	113 (57.1%)	124 (62.3%)	131 (63.9%)	480 (59.6%)
Male	85 (41.9%)	84 (42.4%)	68 (34.2%)	70 (34.1%)	307 (38.1%)
Non-binary	4 (2.0%)	1 (0.5%)	7 (3.5%)	3 (1.5%)	15 (1.9%)
Unknown	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)
Missing	2 (1.0%)	0 (0.0%)	0 (0.0%)	1 (0.5%)	3 (0.4%)
Ethnicity, n (%)
Hispanic or Latino	24 (11.8%)	13 (6.6%)	17 (8.5%)	18 (8.8%)	72 (8.9%)
Not Hispanic or Latino	172 (84.7%)	182 (91.9%)	178 (89.4%)	187 (91.2%)	719 (89.3%)
Unknown/Not reported	7 (3.4%)	3 (1.5%)	4 (2.0%)	0 (0.0%)	14 (1.7%)
Missing	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)
Race, n (%)
White	142 (70.0%)	144 (72.7%)	144 (72.4%)	142 (69.3%)	572 (71.1%)
Black, African American, or Black Multiracial	24 (11.8%)	21 (10.6%)	25 (12.6%)	31 (15.1%)	101 (12.5%)
Asian or Asian Multiracial	26 (12.8%)	22 (11.1%)	18 (9.0%)	23 (11.2%)	89 (11.1%)
Indigenous or Indigenous Multiracial	1 (0.5%)	4 (2.0%)	2 (1.0%)	5 (2.4%)	12 (1.5%)
Unknown or Not Reported	9 (4.4%)	7 (3.5%)	10 (5.0%)	4 (2.0%)	30 (3.7%)
Missing	1 (0.5%)	0 (0.0%)	0 (0.0%)	0 (0.0%)	1 (0.1%)
Highest level of education completed, n (%)
Did not complete Secondary School or High School	6 (3.0%)	6 (3.0%)	8 (4.0%)	8 (3.9%)	28 (3.5%)
High School or Secondary School Degree Complete	31 (15.3%)	24 (12.1%)	33 (16.6%)	41 (20.0%)	129 (16.0%)
Associate's or Technical Degree Complete	33 (16.3%)	27 (13.6%)	36 (18.1%)	35 (17.1%)	131 (16.3%)
College or Baccalaureate Degree Complete	75 (36.9%)	85 (42.9%)	65 (32.7%)	69 (33.7%)	294 (36.5%)
Doctoral or Postgraduate Education	56 (27.6%)	56 (28.3%)	57 (28.6%)	51 (24.9%)	220 (27.3%)
Missing	2 (1.0%)	0 (0.0%)	0 (0.0%)	1 (0.5%)	3 (0.4%)
Current employment status, n (%)
Full-time employment	102 (50.2%)	101 (51.0%)	112 (56.3%)	84 (41.0%)	399 (49.6%)
Not employedc	64 (31.5%)	72 (36.4%)	60 (30.2%)	83 (40.5%)	279 (34.7%)
Part-time employment	35 (17.2%)	25 (12.6%)	27 (13.6%)	37 (18.0%)	124 (15.4%)
Missing	2 (1.0%)	0 (0.0%)	0 (0.0%)	1 (0.5%)	3 (0.4%)
Current relationship status, n (%)
Married	97 (47.8%)	117 (59.1%)	94 (47.2%)	99 (48.3%)	407 (50.6%)
Never married	53 (26.1%)	37 (18.7%)	49 (24.6%)	46 (22.4%)	185 (23.0%)
Domestic partner	13 (6.4%)	12 (6.1%)	17 (8.5%)	10 (4.9%)	52 (6.5%)
Widowed	6 (3.0%)	11 (5.6%)	6 (3.0%)	7 (3.4%)	30 (3.7%)
Divorced or separated	32 (15.8%)	21 (10.6%)	33 (16.6%)	42 (20.5%)	128 (15.9%)
Missing	2 (1.0%)	0 (0.0%)	0 (0.0%)	1 (0.5%)	3 (0.4%)
Number of people living in household, n (%)
1	50 (24.6%)	43 (21.7%)	53 (26.6%)	58 (28.3%)	204 (25.3%)
2	85 (41.9%)	88 (44.4%)	62 (31.2%)	87 (42.4%)	322 (40.0%)
3	30 (14.8%)	25 (12.6%)	40 (20.1%)	19 (9.3%)	114 (14.2%)
4	25 (12.3%)	29 (14.6%)	29 (14.6%)	23 (11.2%)	106 (13.2%)
5	8 (3.9%)	9 (4.5%)	8 (4.0%)	13 (6.3%)	38 (4.7%)
6	3 (1.5%)	3 (1.5%)	7 (3.5%)	1 (0.5%)	14 (1.7%)
7	0 (0.0%)	1 (0.5%)	0 (0.0%)	1 (0.5%)	2 (0.2%)
8	0 (0.0%)	0 (0.0%)	0 (0.0%)	2 (1.0%)	2 (0.2%)
Missing	2 (1.0%)	0 (0.0%)	0 (0.0%)	1 (0.5%)	3 (0.4%)
Annual household income from all sources, n (%)
Less than $10 000	11 (5.4%)	4 (2.0%)	9 (4.5%)	12 (5.9%)	36 (4.5%)
$10 000 to $24 999	20 (9.9%)	14 (7.1%)	16 (8.0%)	23 (11.2%)	73 (9.1%)
$25 000 to $34 999	15 (7.4%)	11 (5.6%)	15 (7.5%)	16 (7.8%)	57 (7.1%)
$35 000 to $49 999	15 (7.4%)	15 (7.6%)	21 (10.6%)	13 (6.3%)	64 (8.0%)
$50 000 to $74 999	24 (11.8%)	21 (10.6%)	25 (12.6%)	20 (9.8%)	90 (11.2%)
$75 000 to $99 999	21 (10.3%)	19 (9.6%)	22 (11.1%)	28 (13.7%)	90 (11.2%)
$100 000 to $149 999	31 (15.3%)	24 (12.1%)	34 (17.1%)	39 (19.0%)	128 (15.9%)
$150 000 to $199 999	17 (8.4%)	21 (10.6%)	14 (7.0%)	6 (2.9%)	58 (7.2%)
$200 000 or more	22 (10.8%)	39 (19.7%)	20 (10.1%)	20 (9.8%)	101 (12.5%)
Prefer not to answer	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)
Missing	27 (13.3%)	30 (15.2%)	23 (11.6%)	28 (13.7%)	108 (13.4%)
PEG score (0 to 10)	5.4 (2.0)	5.1 (1.9)	5.3 (1.8)	5.5 (1.9)	5.3 (1.9)
PEG score, median (range)	5.3 (0.7, 10.0)	5.0 (0.7, 9.7)	5.3 (0.7, 10.0)	5.3 (1.0, 9.7)	5.3 (0.7, 10.0)
Self-reported low back pain duration (months)	163.4 (150.0)	156.7 (151.0)	156.4 (139.8)	154.7 (123.4)	157.8 (141.2)
Self-reported low back pain duration, n (%)
< 1 year	12 (5.9%)	14 (7.1%)	9 (4.5%)	15 (7.3%)	50 (6.2%)
1-5 years	70 (34.5%)	66 (33.3%)	71 (35.7%)	67 (32.7%)	274 (34.0%)
> 5 years	121 (59.6%)	118 (59.6%)	119 (59.8%)	123 (60.0%)	481 (59.8%)
Missing	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)
Low back pain frequency in the past 6 months, n (%)
Every day or nearly every day	181 (89.2%)	163 (82.3%)	160 (80.4%)	169 (82.4%)	673 (83.6%)
At least half the days	22 (10.8%)	35 (17.7%)	39 (19.6%)	36 (17.6%)	132 (16.4%)
Low back pain specific pain intensity, (0 to 10)	5.6 (1.9)	5.4 (1.8)	5.6 (1.7)	5.8 (1.7)	5.6 (1.8)
Low back pain SPI, median (range)	5.0 (1.0, 10.0)	6.0 (1.0, 10.0)	6.0 (1.0, 10.0)	6.0 (1.0, 9.0)	6.0 (1.0, 10.0)
Ever had low back operation, n (%)
No	176 (86.7%)	174 (87.9%)	177 (88.9%)	171 (83.4%)	698 (86.7%)
Yes, at least one	25 (12.3%)	24 (12.1%)	22 (11.1%)	33 (16.1%)	104 (12.9%)
Missing	2 (1.0%)	0 (0.0%)	0 (0.0%)	1 (0.5%)	3 (0.4%)
When was last back operation, n (%) d
Less than 6 months ago	20 (9.9%)	18 (9.1%)	15 (7.5%)	21 (10.2%)	74 (9.2%)
More than 6 months ago but less than 1 year ago	5 (2.5%)	6 (3.0%)	7 (3.5%)	12 (5.9%)	30 (3.7%)
Between 1 and 2 years ago	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)
More than 2 years ago	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)
Any back operations involve a spinal fusion, n (%) d
No	15 (7.4%)	13 (6.6%)	13 (6.5%)	20 (9.8%)	61 (7.6%)
Yes	7 (3.4%)	9 (4.5%)	7 (3.5%)	10 (4.9%)	33 (4.1%)
Not sure	3 (1.5%)	2 (1.0%)	2 (1.0%)	3 (1.5%)	10 (1.2%)
Ever unemployed for 1 or more months due to low back pain, n (%)
No	145 (71.4%)	151 (76.3%)	142 (71.4%)	134 (65.4%)	572 (71.1%)
Yes	40 (19.7%)	27 (13.6%)	36 (18.1%)	43 (21.0%)	146 (18.1%)
Does not apply	16 (7.9%)	20 (10.1%)	21 (10.6%)	27 (13.2%)	84 (10.4%)
Missing	2 (1.0%)	0 (0.0%)	0 (0.0%)	1 (0.5%)	3 (0.4%)
Ever filed or awarded worker's compensation claim related to back problem, n (%)
No	179 (88.2%)	176 (88.9%)	175 (87.9%)	174 (84.9%)	704 (87.5%)
Yes	6 (3.0%)	9 (4.5%)	7 (3.5%)	7 (3.4%)	29 (3.6%)
Does not apply	16 (7.9%)	13 (6.6%)	17 (8.5%)	23 (11.2%)	69 (8.6%)
Missing	2 (1.0%)	0 (0.0%)	0 (0.0%)	1 (0.5%)	3 (0.4%)
Involved in a lawsuit or legal claim related to back problem, n (%)
No	199 (98.0%)	196 (99.0%)	198 (99.5%)	201 (98.0%)	794 (98.6%)
Yes	0 (0.0%)	1 (0.5%)	0 (0.0%)	1 (0.5%)	2 (0.2%)
Not sure	2 (1.0%)	1 (0.5%)	1 (0.5%)	2 (1.0%)	6 (0.7%)
Missing	2 (1.0%)	0 (0.0%)	0 (0.0%)	1 (0.5%)	3 (0.4%)
Ever applied for, or received, disability insurance for pain condition, n (%)
No	180 (88.7%)	185 (93.4%)	183 (92.0%)	170 (82.9%)	718 (89.2%)
Yes	21 (10.3%)	13 (6.6%)	16 (8.0%)	34 (16.6%)	84 (10.4%)
Missing	2 (1.0%)	0 (0.0%)	0 (0.0%)	1 (0.5%)	3 (0.4%)
BMI	28.8 (6.3)	29.3 (7.2)	29.0 (6.3)	29.4 (6.9)	29.1 (6.7)
Systolic blood pressure (mmHg)	125.8 (15.5)	129.5 (17.2)	124.2 (14.3)	125.6 (13.6)	126.3 (15.3)
Diastolic blood pressure (mmHg)	78.5 (10.5)	78.7 (10.2)	76.7 (9.6)	77.7 (10.1)	77.9 (10.1)
Heart rate (bpm)	71.7 (12.9)	70.9 (11.6)	73.2 (12.0)	72.8 (12.3)	72.2 (12.2)
Ever hip replacement surgery
Yes	7 (3.4%)	7 (3.5%)	9 (4.5%)	9 (4.4%)	32 (4.0%)
No	196 (96.6%)	191 (96.5%)	190 (95.5%)	196 (95.6%)	773 (96.0%)
Missing	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)
Ever knee replacement surgery
Yes	13 (6.4%)	13 (6.6%)	15 (7.5%)	9 (4.4%)	50 (6.2%)
No	190 (93.6%)	185 (93.4%)	184 (92.5%)	196 (95.6%)	755 (93.8%)
Missing	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)
Observed gait
Normal	179 (88.2%)	164 (82.8%)	172 (86.4%)	169 (82.4%)	684 (85.0%)
Antalgic	24 (11.8%)	34 (17.2%)	27 (13.6%)	36 (17.6%)	121 (15.0%)
Missing	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)
Tobacco use in previous 12 months
Daily or almost daily	14 (6.9%)	15 (7.6%)	14 (7.0%)	25 (12.2%)	68 (8.4%)
Less than daily	18 (8.9%)	17 (8.6%)	17 (8.5%)	20 (9.8%)	72 (8.9%)
Never	169 (83.3%)	166 (83.8%)	168 (84.4%)	159 (77.6%)	662 (82.2%)
Missing	2 (1.0%)	0 (0.0%)	0 (0.0%)	1 (0.5%)	3 (0.4%)
Alcohol use in previous 12 months e
At least weekly	15 (7.4%)	12 (6.1%)	13 (6.5%)	11 (5.4%)	51 (6.3%)
Monthly	8 (3.9%)	15 (7.6%)	12 (6.0%)	25 (12.2%)	60 (7.5%)
Less than monthly	52 (25.6%)	52 (26.3%)	52 (26.1%)	46 (22.4%)	202 (25.1%)
Never	125 (61.6%)	118 (59.6%)	122 (61.3%)	122 (59.5%)	487 (60.5%)
Missing	3 (1.5%)	1 (0.5%)	0 (0.0%)	1 (0.5%)	5 (0.6%)
Drug use in previous 12 months
Daily or almost daily	11 (5.4%)	5 (2.5%)	8 (4.0%)	11 (5.4%)	35 (4.3%)
Between monthly to weekly	14 (6.9%)	8 (4.0%)	11 (5.5%)	13 (6.3%)	46 (5.7%)
Less than monthly	11 (5.4%)	17 (8.6%)	13 (6.5%)	22 (10.7%)	63 (7.8%)
Never	165 (81.3%)	168 (84.8%)	167 (83.9%)	158 (77.1%)	658 (81.7%)
Missing	2 (1.0%)	0 (0.0%)	0 (0.0%)	1 (0.5%)	3 (0.4%)
Prescription drug used not as intended in previous 12 months
At least once	13 (6.4%)	16 (8.1%)	12 (6.0%)	19 (9.3%)	60 (7.5%)
Never	188 (92.6%)	182 (91.9%)	187 (94.0%)	185 (90.2%)	742 (92.2%)
Missing	2 (1.0%)	0 (0.0%)	0 (0.0%)	1 (0.5%)	3 (0.4%)
Currently taking opioid medication (daily), n (%)
Yes	11 (5.4%)	8 (4.0%)	9 (4.5%)	13 (6.3%)	41 (5.1%)
No	192 (94.6%)	190 (96.0%)	190 (95.5%)	192 (93.7%)	764 (94.9%)
Not sure	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)
Previously diagnosed with COVID-19
Yes	112 (55.2%)	98 (49.5%)	107 (53.8%)	104 (50.7%)	421 (52.3%)
No	84 (41.4%)	95 (48.0%)	85 (42.7%)	98 (47.8%)	362 (45.0%)
Not sure	4 (2.0%)	5 (2.5%)	5 (2.5%)	1 (0.5%)	15 (1.9%)
Prefer not to answer	3 (1.5%)	0 (0.0%)	2 (1.0%)	2 (1.0%)	7 (0.9%)
Missing	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)
Long COVID
Yes	20 (9.9%)	18 (9.1%)	21 (10.6%)	23 (11.2%)	82 (10.2%)
No	79 (38.9%)	74 (37.4%)	81 (40.7%)	74 (36.1%)	308 (38.3%)
Not sure	13 (6.4%)	6 (3.0%)	5 (2.5%)	7 (3.4%)	31 (3.9%)
Prefer not to answer	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)
No self-reported COVID-19 Diagnosis	91 (44.8%)	100 (50.5%)	92 (46.2%)	101 (49.3%)	384 (47.7%)
COVID-19 vaccine
Yes	186 (91.6%)	188 (94.9%)	182 (91.5%)	189 (92.2%)	745 (92.5%)
No	15 (7.4%)	8 (4.0%)	17 (8.5%)	14 (6.8%)	54 (6.7%)
Not sure	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)
Prefer not to answer	2 (1.0%)	2 (1.0%)	0 (0.0%)	2 (1.0%)	6 (0.7%)
Fear-avoidance beliefs about Physical Activity Raw Scoring Scale 2	13.7 (5.1)	12.3 (5.4)	13.3 (5.2)	13.4 (5.6)	13.2 (5.3)
GAD-2 Raw Score	1.7 (1.8)	1.1 (1.2)	1.5 (1.7)	1.4 (1.6)	1.4 (1.6)
Keele STarT Back Screening Tool Risk, n (%)
Low risk	72 (35.5%)	92 (46.5%)	73 (36.7%)	79 (38.5%)	316 (39.3%)
Medium risk	100 (49.3%)	79 (39.9%)	80 (40.2%)	86 (42.0%)	345 (42.9%)
High risk	31 (15.3%)	27 (13.6%)	44 (22.1%)	40 (19.5%)	142 (17.6%)
Missing	0 (0.0%)	0 (0.0%)	2 (1.0%)	0 (0.0%)	2 (0.2%)
Oswestry Disability Index Percentage	29.4 (13.5)	27.7 (12.0)	30.7 (14.1)	31.1 (14.3)	29.8 (13.6)
ODI Percentage, median (range)	28.0 (4.0, 86.7)	26.7 (4.4, 70.0)	28.0 (6.0, 72.0)	28.9 (4.0, 78.0)	28.0 (4.0, 86.7)
Pain Catastrophizing Scale SF-6	11.6 (5.3)	10.4 (4.5)	10.9 (4.9)	11.7 (5.0)	11.1 (5.0)
PainDetect Questionnaire Raw Score
Nociceptive pain	136 (67.0%)	154 (77.8%)	142 (71.4%)	147 (71.7%)	579 (71.9%)
Possible neuropathic pain	38 (18.7%)	33 (16.7%)	35 (17.6%)	36 (17.6%)	142 (17.6%)
Neuropathic pain	28 (13.8%)	11 (5.6%)	20 (10.1%)	20 (9.8%)	79 (9.8%)
Missing	1 (0.5%)	0 (0.0%)	2 (1.0%)	2 (1.0%)	5 (0.6%)
PHQ-2 Raw Score	1.6 (1.6)	1.2 (1.2)	1.4 (1.6)	1.6 (1.6)	1.5 (1.5)
HEAL Positive Outlook Raw Score	21.6 (5.6)	23.0 (4.9)	21.6 (5.8)	21.2 (5.4)	21.8 (5.4)
PROMIS-Cognitive Function—Abilities 2a Raw Score	7.2 (1.8)	7.5 (1.6)	7.1 (1.8)	7.2 (1.8)	7.3 (1.8)
General Sensory Sensitivity Score—External	0.9 (1.4)	0.7 (1.2)	1.1 (1.5)	0.9 (1.3)	0.9 (1.3)
General Sensory Sensitivity Score—Interoception	0.6 (0.8)	0.4 (0.7)	0.6 (0.8)	0.6 (0.8)	0.6 (0.8)
General Sensory Sensitivity Score—Total	1.5 (1.8)	1.1 (1.6)	1.7 (2.0)	1.5 (1.8)	1.5 (1.8)
Self-reported sleep duration in the past month (hours)	6.6 (1.2)	6.7 (1.5)	6.5 (1.2)	6.6 (1.6)	6.6 (1.4)
Symptom Severity Index	5.5 (2.9)	4.8 (2.4)	5.5 (2.9)	5.9 (2.7)	5.5 (2.8)
Widespread Pain Raw Score	2.3 (1.8)	2.1 (1.8)	2.4 (1.9)	2.5 (1.7)	2.3 (1.8)
Stomach pain in the past 4 weeks
Not bothered at all	111 (54.7%)	120 (60.6%)	106 (53.3%)	109 (53.2%)	446 (55.4%)
Bothered a little	71 (35.0%)	66 (33.3%)	78 (39.2%)	77 (37.6%)	292 (36.3%)
Bothered a lot	19 (9.4%)	12 (6.1%)	15 (7.5%)	18 (8.8%)	64 (8.0%)
Missing	2 (1.0%)	0 (0.0%)	0 (0.0%)	1 (0.5%)	3 (0.4%)
Headaches in the past 4 weeks
Not bothered at all	86 (42.4%)	90 (45.5%)	85 (42.7%)	77 (37.6%)	338 (42.0%)
Bothered a little	95 (46.8%)	96 (48.5%)	92 (46.2%)	103 (50.2%)	386 (48.0%)
Bothered a lot	20 (9.9%)	12 (6.1%)	22 (11.1%)	24 (11.7%)	78 (9.7%)
Missing	2 (1.0%)	0 (0.0%)	0 (0.0%)	1 (0.5%)	3 (0.4%)
Radiating pain to buttock/thigh, past 2 weeks
Yes	117 (57.6%)	101 (51.0%)	121 (60.8%)	132 (64.4%)	471 (58.5%)
No	65 (32.0%)	77 (38.9%)	66 (33.2%)	62 (30.2%)	270 (33.5%)
Not sure	19 (9.4%)	20 (10.1%)	12 (6.0%)	10 (4.9%)	61 (7.6%)
Missing	2 (1.0%)	0 (0.0%)	0 (0.0%)	1 (0.5%)	3 (0.4%)
Radiating pain to below knee, past 2 weeks
Yes	53 (26.1%)	40 (20.2%)	52 (26.1%)	56 (27.3%)	201 (25.0%)
No	126 (62.1%)	143 (72.2%)	136 (68.3%)	132 (64.4%)	537 (66.7%)
Not sure	22 (10.8%)	15 (7.6%)	11 (5.5%)	16 (7.8%)	64 (8.0%)
Missing	2 (1.0%)	0 (0.0%)	0 (0.0%)	1 (0.5%)	3 (0.4%)
Social Determinants of Health: Transportation Needs
Yes	17 (8.4%)	13 (6.6%)	14 (7.0%)	18 (8.8%)	62 (7.7%)
No	184 (90.6%)	185 (93.4%)	185 (93.0%)	186 (90.7%)	740 (91.9%)
Missing	2 (1.0%)	0 (0.0%)	0 (0.0%)	1 (0.5%)	3 (0.4%)
Social Determinants of Health: Healthcare Needs
Yes	33 (16.3%)	28 (14.1%)	37 (18.6%)	31 (15.1%)	129 (16.0%)
No	168 (82.8%)	170 (85.9%)	162 (81.4%)	173 (84.4%)	673 (83.6%)
Missing	2 (1.0%)	0 (0.0%)	0 (0.0%)	1 (0.5%)	3 (0.4%)
Social Determinants of Health: Food Insecurity
Often true	7 (3.4%)	3 (1.5%)	7 (3.5%)	7 (3.4%)	24 (3.0%)
Sometimes true	27 (13.3%)	17 (8.6%)	24 (12.1%)	32 (15.6%)	100 (12.4%)
Never true	167 (82.3%)	178 (89.9%)	168 (84.4%)	165 (80.5%)	678 (84.2%)
Missing	2 (1.0%)	0 (0.0%)	0 (0.0%)	1 (0.5%)	3 (0.4%)
Social Determinants of Health: Food Money
Often true	5 (2.5%)	4 (2.0%)	7 (3.5%)	6 (2.9%)	22 (2.7%)
Sometimes true	20 (9.9%)	9 (4.5%)	21 (10.6%)	22 (10.7%)	72 (8.9%)
Never true	176 (86.7%)	185 (93.4%)	171 (85.9%)	176 (85.9%)	708 (88.0%)
Missing	2 (1.0%)	0 (0.0%)	0 (0.0%)	1 (0.5%)	3 (0.4%)
Social Determinants of Health: Utilities
Yes	5 (2.5%)	9 (4.5%)	6 (3.0%)	11 (5.4%)	31 (3.9%)
No	196 (96.6%)	189 (95.5%)	193 (97.0%)	193 (94.1%)	771 (95.8%)
Missing	2 (1.0%)	0 (0.0%)	0 (0.0%)	1 (0.5%)	3 (0.4%)
Social Determinants of Health: Stable Housing
Yes	14 (6.9%)	6 (3.0%)	12 (6.0%)	16 (7.8%)	48 (6.0%)
No	187 (92.1%)	192 (97.0%)	187 (94.0%)	188 (91.7%)	754 (93.7%)
Missing	2 (1.0%)	0 (0.0%)	0 (0.0%)	1 (0.5%)	3 (0.4%)
Social Determinants of Health: Emotional Support
Yes	185 (91.1%)	180 (90.9%)	184 (92.5%)	187 (91.2%)	736 (91.4%)
No	16 (7.9%)	17 (8.6%)	15 (7.5%)	17 (8.3%)	65 (8.1%)
Missing	2 (1.0%)	1 (0.5%)	0 (0.0%)	1 (0.5%)	4 (0.5%)
Social Determinants of Health: Number of Close Friends
0	1 (0.5%)	8 (4.0%)	6 (3.0%)	2 (1.0%)	17 (2.1%)
1-2	46 (22.7%)	41 (20.7%)	46 (23.1%)	49 (23.9%)	182 (22.6%)
3-5	99 (48.8%)	89 (44.9%)	90 (45.2%)	94 (45.9%)	372 (46.2%)
6-10	36 (17.7%)	40 (20.2%)	40 (20.1%)	48 (23.4%)	164 (20.4%)
More than 10	19 (9.4%)	20 (10.1%)	17 (8.5%)	11 (5.4%)	67 (8.3%)
Missing	2 (1.0%)	0 (0.0%)	0 (0.0%)	1 (0.5%)	3 (0.4%)
Perceived Discrimination: Race/Ethnicity
Never	113 (55.7%)	120 (60.6%)	115 (57.8%)	126 (61.5%)	474 (58.9%)
Rarely	47 (23.2%)	42 (21.2%)	45 (22.6%)	40 (19.5%)	174 (21.6%)
Sometimes	40 (19.7%)	36 (18.2%)	39 (19.6%)	38 (18.5%)	153 (19.0%)
Often to almost always	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)
Not sure	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)
Missing	3 (1.5%)	0 (0.0%)	0 (0.0%)	1 (0.5%)	4 (0.5%)
Perceived Discrimination: Orientation/Gender Identity
Never	161 (79.3%)	172 (86.9%)	159 (79.9%)	166 (81.0%)	658 (81.7%)
Rarely	24 (11.8%)	16 (8.1%)	20 (10.1%)	19 (9.3%)	79 (9.8%)
Sometimes	16 (7.9%)	10 (5.1%)	20 (10.1%)	19 (9.3%)	65 (8.1%)
Often to almost always	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)
Not sure	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)
Missing	2 (1.0%)	0 (0.0%)	0 (0.0%)	1 (0.5%)	3 (0.4%)

^aDiscrete variable rows display count and percent, N (%). Continuous variable rows display mean and standard deviation, XX.X (YY.Y).

^b“Sex assigned at birth” is defined as the most commonly reported identity collected from a participant across all visits.

^cCounts retired, unemployed, and other participants who are not currently in the labor force.

^dCounts only include respondents who have answered either “Yes, one operation” or “Yes, more than one operation” when asked about prior low-back operations.

^eParticipants who self-reported male sex at birth were asked about the frequency of >4 alcoholic drinks, while participants who self-reported female sex at birth were asked about the frequency of >3 alcoholic drinks.

Figure 3.

Visual characteristics of randomized participants. (A) Michigan Body Map results for BEST Trial randomized participants (n=801). 96.4% self-reported chronic low back pain in this measure, with significant buttock and neck pain. (B) Histogram of Michigan Body Map body zones in which at least one region was indicated by a participant as having chronic pain. The 7 zones are defined as left arm, right arm, left leg, right leg, chest, back, and head with neck. (C) PROMIS T-scores for BEST Trial randomized participants at week 0. PROMIS scores are scaled such that the population average of the construct is 50. The ridge plots represent density estimates of the distributions of each of the PROMIS constructs among participants who had scorable measurements. These illustrate the distributions of the PROMIS scores in the BEST Trial cohort at baseline. Participants had higher than average levels of pain interference with daily life, fatigue, and disturbances to sleep, and lower than average levels of social participation and physical function. Anxiety and depression in this cohort were bimodal with a proportion of participants with no self-reported depression or anxiety symptoms, and some experiencing significant anxiety and depression. The first row of text values in while boxes above the ridge plots represent the mean and standard deviation PROMIS T-score for each measurement. The percentages in the white boxes are the proportion of participants that were not satisfied with their current level of PROMIS score. (D) After ascertaining PROMIS scores for each secondary study outcome at baseline, participants were asked which areas were most important to them to improve with respect to limitations caused by pain. Participants could report up to 3 answers. The combination of physical function, pain interference, and pain intensity were the most common set of outcomes that BEST Trial participants wanted to improve upon.

Phenotyping assessments

BEST collected rich phenotyping assessments at the baseline study visit of cLBP participants beginning treatment, which consisted of 7 distinct assessments and 3 to 4 hours of participant time. Participants could opt in to 3 to 5 additional hours of assessments of optional phenotyping.¹³ 70.6% of randomized participants completed all required assessments except for the basic spine MRI and patient preference survey. Additionally, 346 (43.0%) randomized participants completed all required phenotyping assessments, including PROs (100.0% have at least 1 questionnaire complete), stool collection (98.8%), food frequency questionnaire (96.3%), biomechanical assessment (91.8%), biospecimen collection (88.6%), patient preferences survey (80.3%), and basic spine MRI (75.0%) (Figure S7). In addition to the required phenotyping, 510 participants (63.3%) received at least 1 of the 9 optional phenotyping assessments. The optional phenotyping assessments have been described elsewhere.¹³ Of all the participants that completed at least 1 optional phenotypic assessment, 55 completed all 9 optional assessments (10.7%), and 44 completed all but the advanced spine MRI (8.6%) (Figure S8).

Discussion

The randomized participants in the BEST Trial approximately matched the distribution of race, biological sex, and employment observed in epidemiological studies of cLBP patients in the United States.¹ Additionally, the study sample approximately matches the observed distributions for age and sex of other clinical trials of cLBP in the United States and Europe.^42–46 The BEST Trial had more highly educated participants than previous observational studies, but this observation is consistent with other clinical trials.⁴⁷

Our success in recruiting a cohort of randomized participants that matched the racial distribution of chronic back pain in the United States was driven by site-level efforts and supplemental NIH funds targeting under-represented racial subgroups in clinical trials. Study teams performed outreach to community programs and attended events to facilitate the engagement of under-represented groups. These activities were supported by additional coordinator effort and recruiting costs, provided to sites through a Diversity Supplement from NIAMS (3U24AR076730-01S1). However, more can still be done to recruit underrepresented racial minority groups in clinical trials due to their disproportionate burden of cLBP.⁴⁸

In addition to demographic features, PRO measures are consistent with those of a population experiencing cLBP and seeking care.^1,45 From comprehensive patient reported pain outcomes, including the Michigan Body Map, we confirmed that we enrolled a sample of participants with chronic pain, where low back pain is their primary source of pain. The mean duration of cLBP was 163.4 months (13.6 years), and 40% of the participants self-reported cLBP of <5 years. This significant variability is confirmed by a standard deviation of 150.0 months (12.5 years). Compared to normative populations, as measured by PROMIS measures, the population had a high degree of pain interference, above-average sleep disturbance, fatigue, anxiety, and depression, and slightly below-average social participation and physical function. This is important as the psychosocial context of our patient sample is consistent with the clinical population of patients participating in surveys and clinical trials to treat cLBP.^1,45 Matching on important demographic, pain, and psychosocial characteristics suggests that the effectiveness of biomarker-based treatment assignments will be generalizable to the United States cLBP population.

BEST Trial randomized participants reported that improving physical function, pain intensity, and pain interference were most important to them regarding limitations caused by their pain. Several primary outcomes were considered during the trial's design phase, but the 24-week PEG score was chosen. The domains that are most important to participants are captured by the PEG score, confirming our choice of primary outcome as patient-centered and clinically relevant to the trial population. There was only a relatively small number of participants who reported that improving their anxiety and depression was important to them.¹³

Future trials in cLBP participants can build off the experiences with recruitment of the BEST Trial, where we observed that a run-in period with stringent inclusion criteria before randomization contributed to excluding a higher rate of consented participants than initially planned. One originally planned eligibility criterion was engagement with daily pain questionnaires as a proxy to measure overall engagement in the trial. While initially well intentioned, digital engagement of the sort measured in the run-in period was not a part of ongoing follow-up, and thus it became non-essential to exclude participants from the trial based on not answering these daily pain questionnaires. Another lesson learned was the original wording of an eligibility criterion that participants must watch the educational modules before the randomization visit. This resulted in excluding a few highly engaged contented participants because they forgot to watch the videos before the visit. Operationally, this was checked by seeing if the module videos were watched completely, and several participants were excluded who did not watch the last several seconds of the study module. We adjusted this criterion to permit participants to watch the modules at the study visit to allow coordinators to help them troubleshoot the modules and ensure that all participants were exposed to the valuable educational material about the trial without impacting their ability to receive treatment.

The results from our exploratory run-in analysis help characterize the impact of the run-in period on our study sample. In the BEST Trial, 86.3% of patients consented were eligible after the run-in period. Our random forest exploratory analysis of predictors of run-in failure demonstrated that income (range of model-predicted probabilities of eligibility after run-in: 0.85-0.88), cognitive function (0.83-0.875), and depression (0.82-0.87) were most important in predicting run-in eligibility. Our findings suggest both the highest and the lowest income participants were less likely to complete the run-in period (0.86) compared to moderate income participants (0.88). While cognitive function and depression contributing to predicted run-in failure may suggest that participants with stronger pain-related symptoms were more likely to fail run in, PEG score at the pre-screening visit was not useful to predict eventual run-in status. Additionally, the number of daily pain questionnaires required for run-in (accounting for the changing protocol over time) was not predictive of run-in failure, though more work will be done to see if it was associated with future study dropout; it may be true that the lower DPQ requirement led to future increased dropout during the study. One limitation of the model is the lower out-of-sample specificity, which may be due to various reasons that participants fail the run-in period. Factors predicting run-in failure will be assessed in the randomized population to see if these factors negatively impacted study retention. Despite unexpected challenges with the run-in period, the BEST Trial fully recruited its sample size goals.

Both the design and implementation of the BEST Trial are innovative and represent, to our knowledge, one of the largest SMART trials to be performed in a clinical setting for a pain condition. The overarching goal of BEST is to determine if a set of individual characteristics can be identified that improve treatment selection for individuals with cLBP compared to usual practice. This work presents a broad range of patient features observed at baseline and many completed data collection procedures. The participant data from the BEST Trial will enable rich discoveries for the treatment of cLBP, and these data will have significant generalizability to the population of cLBP patients.

Supplementary material

Supplementary material is available at Pain Medicine online.

Funding

The Back Pain Consortium (BACPAC) Research Program is administered by the National Institute of Arthritis and Musculoskeletal and Skin Diseases (NIAMS). This research was supported by the National Institutes of Health through the NIH HEAL Initiative under award number 1U24AR076730. The content is solely the authors' responsibility and does not necessarily represent the official views of the National Institutes of Health or its NIH HEAL Initiative.

Conflicts of interest: B.R., K.S.B., K.M.B., A.C., S.Ma., K.V.P., S.R.P., M.J.S., F.A., B.A., S.B., A.B., J.J.B., A.E.B., T.S.C., J.P.C., L.C., B.C., D.D., J.L.E., S.M.E., A.J.F., A.F., I.G., R.W.G., C.M.G., S.G., C.A.G., L.H., R.E.H., S.E.H., K.H., A.H., A.I., S.J.B., C.K., K.M.K., G.G.K., M.R.K., D.L., R.L., J.L., J.C.L., P.M., J.M., L.M.M., M.McC., M.McM., W.M., R.M., U.H.M., C.O., S.Pea., S.Pel., S.R., A.S., E.S., J.S., N.V., A.M.W., T.E.W., K.A.W., L.W., D.A.W., L.W., J.E.W., F.Z., B.Zha., B.Zho., K.J.A., D.J.C., G.A.S., and M.C.M. do not have conflicts of interest to report. A.K.B. receives funding from the National Institutes of Health for research (1R01AG082777-01A1 and 1R25DA061740-01). A.L.C. receives research funding from the National Institutes of Health (R01NS135833-01A1, RM1NS128956-01A1, R01DA058694-01), the U.S. Department of Defense (W81XWH-21-PRMRP-CTA (PR212158)), and the Cystic Fibrosis Foundation (FRIEDM22A0-I). A.L.C. has also consulted for Swing Therapeutics and Scilex Pharmaceuticals. R.W.H. receives research funding from the National Institutes of Health (U24DA058606, R25DA061740, UH3AR077360) and from Nevro, Inc. R.W.H. has also consulted for State Farm, Inc. J.S.D. is the CEO and a shareholder of Conity, Inc. Conity, Inc. was recently formed and is seeking a license from The Ohio State University (OSU) to commercialize the OSU technology used in this study. J.M.F. receives research funding from the National Institutes of Health, the Patient-Centered Outcomes Research Institute, the Department of Defense and the Centers for Disease Control and Prevention. A.L.H. has consulted for Community Health Focus, Inc. G.K. has consulted for Bioniks. S. Majumdar receives research funding from GE Healthcare and Siemens Healthineers. W.S.M. recently formed Conity, Inc. Conity, Inc. is seeking a license from The Ohio State University (OSU) to commercialize the OSU technology used in this study. S.A.M. has served as a consultant for Walter Reed Army Institute for Research, Arbor Medical Innovations, and BioXcel Therapeutics, Inc. V.N. is a paid consultant for Cala Health, a bioelectronic medicine company developing wearable neuromodulation therapies. S.D.R. consulted for Medbridge Inc. in 2023. M.S.W. has consulted for TerSera Therapeutics. A.D.W. has consulted for Vertex Pharmaceuticals and Seikagaku Corporation North America.