Biopsychosocial influence on shoulder pain: results from a randomized pre-clinical trial of exercise-induced muscle injury

INTRODUCTION

Chronic musculoskeletal pain conditions are among the most prevalent, disabling, and costly medical problems experienced by individuals globally. Preventing chronic pain is essential for improving patient care. Pain relief that is matched to specific risk characteristics (ie, personalized or precision medicine) is a high priority for future research and practice initiatives.[1; 25] Personalized medicine options exist for cardiac medicine[37; 38] and oncology[3; 11], but few models are available for pain management.[31]

Chronic musculoskeletal pain has a complex biopsychosocial etiology, which is one likely reason for the lack of personalized medicine options for pain management.[21] While cardiac medicine and oncology have focused on the identification of genetic factors, it has been posited that personalized medicine options for pain will require genetic factors in combination with psychological, environmental, and/or social risk factors.[13; 31] Indeed, our Biopsychosocial Influence on Shoulder Pain (BISP) line of research has implemented this multiple risk factor approach and successfully validated a high-risk subgroup comprised of psychological and genetic factors.[20] One component of this high-risk subgroup, the catechol-O-methyltransferase (COMT) gene, encodes the COMT enzyme, which metabolizes catecholamines. COMT polymorphisms and haplotypes associated with low COMT activity have been linked to pain sensitivity and an increased risk of multiple musculoskeletal pain conditions.[5; 8; 29; 30; 41] Pain catastrophizing, the psychological component of this high-risk subgroup, is comprised of pain-related rumination, magnification, and thoughts of helplessness/pessimism. Pain catastrophizing has a well-established link to pain perception and self-reported disability in multiple pain populations.[12; 14; 44]

This high-risk subgroup has been exclusively investigated in predictive cohort studies of exercise-induced shoulder injury (ie, pre-clinical models) and rotator cuff repair (ie, clinical models). In earlier studies, we reported that 1) an interaction between the COMT genotype and pain catastrophizing was a stronger predictor of shoulder pain and disability outcomes compared with either factor alone[17] and 2) the high-risk subgroup was the only phenotype predictive of higher pain intensity in pre-clinical and clinical cohorts.[20] These robust predictive findings across multiple cohorts provided the impetus to transition our BISP line of research to a randomized clinical trial. Testing interventions matched to the characteristics of the high-risk subgroup would determine the efficacy of personalized treatment options for shoulder pain. Specifically, we targeted catastrophizing with a brief cognitive-behavioral pain intervention, and we targeted the COMT genotype using the beta-adrenergic receptor antagonist propranolol. Furthermore, we used our established exercise muscle injury model to induce shoulder pain. The advantages of this model over the clinical model of rotator cuff repair included a standardized injury model, high treatment fidelity, close monitoring of participant responses on primary and secondary measures, and an established translational link to a post-operative clinical model.[20] Therefore, the purpose of this article is to describe the primary outcomes and findings from the BISP clinical trial.

METHODS

Overview

The BISP clinical trial was designed with reporting following CONSORT[32; 33] and SPIRIT[10] recommendations. This trial was prospectively registered (https://clinicaltrials.gov) as NCT02620579 with a start date of December 3, 2015 and an end date of November 24, 2021. Our published protocol article provides more detail on the study methods.[19] This trial was approved by the University of Florida Institutional Review Board, and all participants provided informed consent for eligibility screening and trial enrollment. Briefly, eligible and enrolled participants had shoulder pain induced via exercise-induced muscle injury (day 1), followed by randomly assigned treatment delivered for 4 consecutive days through onsite visits (days 2 - 5). The four intervention groups were created by crossing two pharmacologic conditions (propranolol or placebo) with two informational conditions (psychologic intervention or general education). The primary statistical analysis outcome was to determine whether the combined personalized intervention group (propranolol with psychologic intervention) reported greater recovery from exercise-induced shoulder pain.

Participants

Healthy participants were considered for study participation. The inclusion criteria were 1) ages ≥ 18 years to 65 years and 2) English-speaking. The exclusion criteria were identified at screening and trial enrollment. All additional exclusion criteria are reported in Table 1 and described in more detail in the subsequent sections.

Table 1.

Exclusion criteria for Biopsychosocial Influence on Shoulder Pain pre-clinical trial

Identified during screening
Exclusion criteria for high-risk subgroup
Pain Catastrophizing Scale Score < 5	COMT rs6269 not AA
Exclusion criteria for exercise-induced shoulder injury protocol
Chronic pain (> 3 months) in any area	Current neck or shoulder pain	Previous history of upper extremity surgery
Neurologic impairment of the upper extremity (determined by loss of sensation, muscle weakness, and reflex change)	Previous history of neck or shoulder pain (operationally defined as experiencing pain longer than 48 hours or seeking medical treatment)	Current or regular use of pain medication
Regular participation in upper extremity weight training
Identified during trial enrollment
Exclusion criteria for propranolol
Clinically significant abnormal 12-lead ECG	Sinus bradycardia (resting heart rate below 55 beats per minute)	Uncontrolled hypertension (resting systolic blood pressure below 90 mm Hg)
Cardiac failure	Coronary heart disease	Wolff-Parkinson-White Syndrome
Greater than first-degree heart block	Known hypersensitivity to propranolol
Identified during screening or trial enrollment
Other medical exclusion criteria for study participation
Bronchial asthma	Nonallergic bronchospasm	History of recent surgery requiring general anesthesia
Diabetes	Pregnancy	Major depression
Chronic obstructive pulmonary disease	Dementia	Breastfeeding

High-Risk Subgroup

In our prior cohort study high-risk subgroup status was determined empirically (ie, with Classification and Regression Tree analysis) from a priori selected genetic and psychological variables [20]. Two factors comprised the high-risk subgroup: 1) COMT AA genotype enriching for increased pain sensitivity at the rs6269 variant and 2) Pain Catastrophizing Scale (PCS) scores of 5 or greater.[20] Those familiar with the PCS will notice that a score of 5 is not considered “elevated” when compared with clinical populations. This cut-off score was used for subgroup classification because it was determined empirically in a pre-clinical cohort and was specific to the COMT rs6269 AA genotype. This high-risk subgroup was derived along with other candidate subgroups; however the subgroup comprised of COMT AA genotype and PCS scores of 5 or greater was the best predictor of increased 12 month post-operative shoulder pain in a clinical cohort. [20] Therefore, it was designated as the high-risk subgroup of interest for a pre-clinical trial.

Screening

Periodic screenings of volunteers were completed on the University of Florida campus and in the local community. All participants provided informed consent to screening, and the criteria were geared towards identifying high-risk group membership and appropriateness of exercise-induced muscle injury (Table 2). Specifically, participants provided a saliva sample for rapid COMT rs6269 genotyping; completed the PCS, a 13-item, 4-point rating scale[43]; and answered questions related to the exercise-induced muscle injury protocol. A small monetary incentive was provided to encourage participation in screening. The rapid rs6269 genotyping was accomplished within 2-4 days of preliminary recruitment, using PCR and restriction digest to help identify the high-risk subjects. PCR primers COMT6269Bst-F (5’-TTCTGAACCTTGCCCCTCTTC) and COMT6269-R (5’-GAACTCGTTCAGCCGATAAGG) were used to generate a 154 bp product, which was digested by BstBI to reveal the genotype. Individuals who participated in the full protocol were also re-genotyped for quality control.

Table 2.

Descriptive summary of the clinical trial participants

Variable	Category	Overall	Group A	Group B	Group C	Group D
Sex	Male	80 (34.3%)	18 (31.6%)	20 (35.1%)	22 (35.5%)	20 (35.1%)
	Female	153 (65.7%)	39 (68.4%)	37 (64.9%)	40 (64.5%)	37 (64.9%)
Dominant Hand	Right	208 (89.3%)	48 (84.2%)	54 (94.7%)	56 (90.3%)	50 (87.7%)
	Left	21 (9.0%)	7 (12.3%)	3 (5.3%)	5 (8.1%)	6 (10.5%)
	Ambidextrous	4 (1.7%)	2 (3.5%)	0 (0.0%)	1 (1.6%)	1 (1.8%)
Age	Mean ± SdtDev	21.0 ± 4.3	21.7 ± 5.9	20.3 ± 2.5	21.4 ± 5.3	20.5 ± 1.9
	Median (Min, Max)	20.0 (18.0, 55.0)	20.0 (18.0, 55.0)	20.0 (18.0, 32.0)	20.0 (18.0, 49.0)	20.0 (18.0, 25.0)
Combined Race	Asian	71 (30.5%)	18 (31.6%)	20 (35.1%)	15 (24.2%)	18 (31.6%)
	White	108 (46.4%)	24 (42.1%)	20 (35.1%)	33 (53.2%)	31 (54.4%)
	Black or Other	54 (23.2%)	15 (26.3%)	17 (29.8%)	14 (22.6%)	8 (14.0%)
Height	Mean ± SdtDev	66.0 ± 4.5	65.7 ± 5.3	66.0 ± 4.2	66.4 ± 4.1	66.0 ± 4.5
	Median (Min, Max)	66.0 (45.0, 78.0)	66.0 (45.0, 78.0)	66.0 (57.0, 77.0)	66.0 (60.0, 76.0)	65.0 (54.0, 74.0)
Weight	Mean ± SdtDev	148.2 ± 34.2	150.0 ± 34.5	148.8 ± 33.7	148.4 ± 36.7	145.4 ± 32.5
	Median (Min, Max)	140.0 (84.0, 290.0)	140.0 (84.0, 250.0)	141.0 (110.0, 260.0)	140.0 (107.0, 290.0)	138.0 (96.0, 248.0)
BMI	Mean ± SdtDev	23.7 ± 4.3	24.1 ± 4.5	23.9 ± 4.1	23.6 ± 4.4	23.4 ± 4.3
	Median (Min, Max)	22.7 (14.6, 42.1)	23.4 (17.0, 41.7)	22.3 (17.9, 39.2)	22.5 (14.6, 42.1)	23.0 (15.3, 37.2)
FPQ Total	Mean ± SdtDev	14.8 ± 6.3	15.6 ± 6.8	14.6 ± 7.0	14.5 ± 6.0	14.5 ± 5.5
	Median (Min, Max)	15.0 (0.0, 31.0)	15.0 (4.0, 28.0)	16.0 (0.0, 31.0)	15.0 (0.0, 25.0)	14.0 (0.0, 26.0)
PCS Total	Mean ± SdtDev	12.3 ± 8.3	12.7 ± 9.0	12.6 ± 7.2	12.6 ± 9.1	11.4 ± 7.8
	Median (Min, Max)	12.0 (0.0, 34.0)	12.0 (0.0, 34.0)	12.0 (0.0, 30.0)	11.0 (0.0, 32.0)	11.0 (0.0, 31.0)
PHQ-9 Total	Mean ± SdtDev	2.7 ± 3.5	3.2 ± 4.4	2.8 ± 3.1	1.9 ± 2.3	2.9 ± 3.8
	Median (Min, Max)	2.0 (0.0, 24.0)	2.0 (0.0, 24.0)	2.0 (0.0, 15.0)	1.0 (0.0, 11.0)	2.0 (0.0, 23.0)
TSK-11 Total	Mean ± SdtDev	19.5 ± 5.1	19.4 ± 5.7	19.5 ± 5.0	19.6 ± 4.7	19.8 ± 5.2
	Median (Min, Max)	19.0 (10.0, 42.0)	19.0 (10.0, 42.0)	19.0 (10.0, 31.0)	20.0 (12.0, 33.0)	20.0 (10.0, 38.0)

Table Key: Group A = Placebo and Psychologic Intervention; Group B = Placebo and General Education; Group C = Propranolol and Psychologic Intervention; Group D = Propranolol and General Education.

BMI, body mass index; FPQ, Fear of Pain Questionnaire; PCS, Pain Catastrophizing Scale; PHQ-9, Patient Health Questionnaire-9; TSK-11, Tampa Scale for Kinesiophobia-11.

Trial Enrollment

Participants who screened as being in the high-risk subgroup and appropriate for exercise-induced muscle injury were eligible for participation in the clinical trial and provided additional informed consent. A separate consent for the clinical trial was required to allow for additional testing to determine the appropriateness of receiving propranolol and also identifying other medical reasons for exclusion (Table 1). Participants who were eligible based on being able to take propranolol and who had no other medical exclusions were then enrolled in the clinical trial.

Exercise-Induced Shoulder Injury

Research personnel performing the muscle injury protocol and follow-up assessments were blinded to randomization. Subjects underwent exercise-induced shoulder injury to the dominant arm. Briefly, the eccentric exercise fatigue protocol used isokinetic equipment and followed an established protocol from prior studies.[7; 17; 18] The goal of the injury protocol was to induce delayed onset muscle soreness in the rotator cuff musculature, resulting in participants experiencing shoulder pain intensity, loss of range of motion, inflammatory responses, and muscle weakness.

Randomization and Allocation

The statistical team (SSW and YD) prepared the randomization scheme by computer and the scheme completed prior to the start of the study. Treatment assignments were accessed by study personnel through a secured website that provided independent assignment for pharmaceutical and psychological interventions. Random assignment was determined in sequential order as each participant entered the study. Research staff were blinded to intervention assignment they were not involved with. For example, the pharmacy staff did not know the assignment for the informational intervention. Randomization was completed prior to the muscle injury protocol and stratified by sex due to reported sex differences in pain conditions[4; 22] and observed sex differences in our prior study of shoulder pain.[28]

Interventions

Personalized Pharmaceutical.

The first pharmaceutical administration occurred before injury, which allows for any immediate pre-emptive effects on inflammatory or pain sensitivity regulation. This administration also matches when propranolol would be administered in the proposed translational clinical model (ie, pre-operatively). The University of Florida Investigational Drug Service prepared long-acting propranolol (Propranolol LA) 60 mg that was administered orally in the Pain Clinical Research Unit once daily. This dose was selected as a bioequivalent dose to those reported in a recent clinical trial examining responses to propranolol among patients with TMD pain.[42] Cardiovascular response was monitored for safety (early identification of potential adverse events) and pharmacologic effect (demonstrate medication absorbed) 60 minutes after drug administration by a research nurse.

Placebo Pharmaceutical.

Placebo capsules prepared by the UF Investigational Drug Service were visually indistinguishable from the active medication. Placebo administration was done in the same fashion as described in the previous Personalized Pharmaceutical section to maintain blinding. This included the same timing for each session and monitoring of cardiovascular responses.

Personalized Psychologic.

The personalized psychologic intervention followed key principles for psychologically informed interventions in musculoskeletal pain and in doing so was intended to induce individual change in beliefs and behavior.[26] These principles are not specific to the anatomical region of pain and involve addressing issues related to pain-related fear, kinesiophobia, and pain catastrophizing. Therefore, our personalized psychological education intervention addressed these factors with special emphasis on pain catastrophizing since that was the primary psychologic factor in defining the high-risk subgroup. The personalized psychologic intervention was administered on days 2-4 following the exercise-induced muscle injury protocol since the intervention was predicated on the individual experiencing pain. The psychologic modules were scripted, and the duration was 10-15 minutes with all content delivered on site. The participants viewed the module on a computer and then interacted with research staff in scripted areas (ie, demonstration of key principle and questions on module content).

General Education.

The general education intervention matched the structure and administration of the personalized intervention. The general education module consisted of content relative to shoulder pain, but the module was not intentionally designed to change thoughts or beliefs about experiencing pain. That is, this intervention was intended to passively disseminate information. The general education intervention modules were also administered days 2-4 following exercise-induced injury with the overall goal of the participant understanding shoulder anatomy and injury. The general education modules were scripted and lasted 10-15 minutes each session. Similar to the personalized intervention, the participants viewed the module on a computer and then interacted with research staff in scripted areas (ie, demonstration of key principle and questions on module content).

Outcomes

Outcomes were selected based on relevance to clinical outcomes and successful use in our prior studies. The Brief Pain Inventory (BPI) was used for pain intensity as it has been found to have good test-rest reliability over short intervals.[27] Using an 11-point numerical rating scale ranging from 0 (no pain) to 10 (worst pain intensity imaginable), participants rated the intensity of current pain and pain intensity at its worst, best, and average over the past 24 hours. Recovery was defined using BPI ratings as a current pain intensity rating of 0/10 and a worst pain intensity rating of less than 2/10. In addition to whether the participant met the recovery criterion (Yes or No), we also recorded the number of days it took to reach this recovery criterion (ie, shoulder pain duration). Additionally, the validated abridged version of the Disabilities of the Arm, Shoulder, and Hand Questionnaire (Quick-DASH) was used as an endpoint to assess upper extremity disability, consisting of 11 functional items, with total scores ranging from 0 (no disability) to 100 (complete disability).[23] Similar to the BPI ratings, Quick-DASH scores were recorded daily until the BPI recovery criterion was met.

Sample Size

Details on our original sample size estimates are provided in the protocol article.[19] Briefly, we projected the need to enroll 448 high-risk subjects to meet our assumptions of 80% power to detect the assumed differences across the four intervention groups and 91% power for the primary comparison between the combined personalized interventions and the combined placebo and general education interventions at a type-I error level of 0.05. Post-study power analysis indicates that the actual sample size (N=119 for propranolol vs N=114 for placebo, N=119 for psychologic intervention vs N=114 for general education) provides 84% power to detect absolute recovery rate difference of 20% (the difference hypothesized in the protocol paper), which corresponds to a Number Needed to Treat =5 for each main effect.

Statistical Analysis Plan

Details for the original statistical plan are provided in the protocol article.[19] All statistical analyses were performed using SAS software, version 9 (SAS Institute Inc, Cary, North Carolina). Summary statistics were provided for baseline measures by intervention groups to assess effects of randomization and to investigate if covariates should be included in the primary analysis. One modification was made to the original statistical plan. That is, there was a planned interim analysis to inform an adaptive randomization scheme that would have potentially enhanced the statistical power for group comparisons of interest. However, we did not reach adequate enrollment numbers; therefore, this interim analysis was not completed. Otherwise, the original statistical plan was followed.

The primary analysis compared recovery rates across the four randomly assigned groups using logistic regression. The primary outcome variable was dichotomized based on meeting the recovery criterion for shoulder pain intensity by at least 6 days. An a priori plan was developed to include age, sex, and race as covariates with additional covariates only considered if there was an imbalance across groups for the variables that correlated with the outcome measures. Any missing outcomes were predicted by subject pain intensity trajectory plus baseline demographic factors. The primary comparison of the combined personalized interventions and the combined placebo and general education condition was tested at the 0.05 significance level. The other between-group differences were tested using Holm’s step-down procedure to ensure that the family-wise error rate was controlled at 0.05.[24]

Given that the responses to these interventions may not occur as hypothesized, we included a plan for additional analyses in the protocol article if no group differences were found in the primary analysis.[19] For example, an analysis was planned to determine if sex-specific intervention effects occurred. We also explored whether self-identified race or individual level of pain catastrophizing differentially affected recovery rates across the four intervention groups. Finally, we investigated via survival analysis whether intervention effects differed based on using an outcome of shoulder pain duration measured by the number of days.

RESULTS

Figure 1 is the CONSORT flow diagram detailing screening and enrollment for the BISP trial. Screening was performed for 1,452 potential candidates, of whom 445 (30.6%) met the requirements for high-risk subgroup membership. Of the 445 candidates eligible for the trial, 181 (40.7%) were excluded for having medical reasons that prevented them for taking propranolol for research purposes, having another medical exclusion, or declining to receive more information on the trial. Therefore, we enrolled and randomized 264 (59.3%) candidates who were eligible. A total of 28 (10.6%) candidates were excluded from the trial post-randomization due to the identification of additional exclusion criteria related to the safety of propranolol administration (eg, abnormal ECG findings) or a candidate not being able to complete the study (eg, change in time commitment). All of these post-randomization exclusions were made for safety reasons and participants were excluded prior to any baseline data collection, induction of exercise-induced muscle injury, and treatment allocation. In addition, no loss to follow-up occurred (n=0) and few participants discontinued the intervention (n=3). Therefore, all participants were analyzed as originally randomized with results presented for completed and predicted outcomes. The predicted outcomes included imputed data for those who discontinued the intervention (n=3).

Figure 1.

CONSORT flow diagram describing screening, enrollment, and follow-up for the BISP trial

BISP, Biopsychosocial Influence on Shoulder Pain.

Table 2 reports descriptive statistics for the 261 participants who were randomized, completed baseline data collection, underwent the exercise-induced muscle injury protocol, and received randomly assigned treatment. The four groups were balanced for baseline variables and accordingly no additional covariates were entered into the logistic regression models. Table 3 reports descriptive statistics for pain intensity and self-reported disability over the 5-day onsite observation period. These descriptive statistics show patterns similar to our prior work, in which rating of the peak worst pain intensity and disability are experienced within 24-48 hours after completing the muscle injury protocol (ie, days 2 and 3). For all participants in the trial, the peak rating for worst pain intensity was 3.0 (SD=2.0) on days 2 and 3. The peak rating for current pain intensity was 1.9 (SD=1.8) on day 2 and 15.0 (SD=13.5) for self-reported disability on day 3.

Table 3.

Descriptive summary of the clinical trial participants

Variable	Time Point		Overall	Group A	Group B	Group C	Group D
Worst Pain Intensity	Day 1	Mean ± SdtDev	0.2 ± 0.7	0.1 ± 0.4	0.2 ± 0.6	0.2 ± 0.6	0.4 ± 1.0
		Median (Min, Max)	0.0 (0.0, 5.0)	0.0 (0.0, 2.0)	0.0 (0.0, 3.0)	0.0 (0.0, 3.0)	0.0 (0.0, 5.0)
	Day 2	Mean ± SdtDev	3.0 ± 2.0	2.9 ± 2.1	3.2 ± 2.1	2.9 ± 1.9	3.1 ± 2.0
		Median (Min, Max)	3.0 (0.0, 8.0)	3.0 (0.0, 8.0)	3.0 (0.0, 8.0)	3.0 (0.0, 7.0)	3.0 (0.0, 8.0)
	Day 3	Mean ± SdtDev	3.0 ± 2.0	2.8 ± 2.2	3.1 ± 2.0	2.8 ± 1.9	3.3 ± 2.0
		Median (Min, Max)	3.0 (0.0, 9.0)	2.0 (0.0, 9.0)	3.0 (0.0, 8.0)	2.0 (0.0, 7.0)	3.0 (0.0, 8.0)
	Day 4	Mean ± SdtDev	2.3 ± 2.0	2.3 ± 2.3	2.3 ± 2.0	2.3 ± 1.9	2.2 ± 1.8
		Median (Min, Max)	2.0 (0.0, 10.0)	2.0 (0.0, 10.0)	2.0 (0.0, 9.0)	2.0 (0.0, 8.0)	2.0 (0.0, 7.0)
	Day 5	Mean ± SdtDev	1.4 ± 1.6	1.6 ± 2.0	1.6 ± 1.9	1.1 ± 1.3	1.3 ± 1.3
		Median (Min, Max)	1.0 (0.0, 8.0)	1.0 (0.0, 7.0)	1.0 (0.0, 8.0)	1.0 (0.0, 5.0)	1.0 (0.0, 6.0)
Current Pain Intensity	Day 1	Mean ± SdtDev	0.1 ± 0.4	0.1 ± 0.2	0.1 ± 0.5	0.1 ± 0.3	0.2 ± 0.6
		Median (Min, Max)	0.0 (0.0, 3.0)	0.0 (0.0, 1.0)	0.0 (0.0, 2.0)	0.0 (0.0, 2.0)	0.0 (0.0, 3.0)
	Day 2	Mean ± SdtDev	1.9 ± 1.8	1.8 ± 1.8	2.1 ± 2.0	1.7 ± 1.6	2.0 ± 1.8
		Median (Min, Max)	1.0 (0.0, 8.0)	1.0 (0.0, 8.0)	2.0 (0.0, 8.0)	1.0 (0.0, 6.0)	2.0 (0.0, 7.0)
	Day 3	Mean ± SdtDev	1.7 ± 1.9	1.6 ± 2.0	1.6 ± 1.8	1.5 ± 1.8	2.0 ± 1.9
		Median (Min, Max)	1.0 (0.0, 9.0)	1.0 (0.0, 9.0)	1.0 (0.0, 7.0)	1.0 (0.0, 8.0)	1.0 (0.0, 8.0)
	Day 4	Mean ± SdtDev	1.1 ± 1.6	1.3 ± 2.0	1.3 ± 1.7	0.9 ± 1.4	0.9 ± 1.3
		Median (Min, Max)	1.0 (0.0, 9.0)	0.5 (0.0, 9.0)	1.0 (0.0, 8.0)	0.5 (0.0, 7.0)	0.0 (0.0, 6.0)
	Day 5	Mean ± SdtDev	0.6 ± 1.5	0.8 ± 1.7	0.7 ± 1.4	0.5 ± 1.8	0.4 ± 0.8
		Median (Min, Max)	0.0 (0.0, 10.0)	0.0 (0.0, 10.0)	0.0 (0.0, 7.0)	0.0 (0.0, 10.0)	0.0 (0.0, 4.0)
Q-Dash Score	Day 1	Mean ± SdtDev	3.8 ± 5.6	4.7 ± 6.5	3.1 ± 4.8	3.7 ± 5.3	3.9 ± 5.7
		Median (Min, Max)	2.3 (0.0, 34.1)	2.3 (0.0, 34.1)	0.0 (0.0, 15.9)	0.0 (0.0, 20.5)	2.3 (0.0, 29.5)
	Day 2	Mean ± SdtDev	13.0 ± 12.0	14.2 ± 12.5	13.2 ± 12.7	11.4 ± 11.5	13.1 ± 11.2
		Median (Min, Max)	9.1 (0.0, 50.0)	9.1 (0.0, 47.7)	9.1 (0.0, 47.7)	9.1 (0.0, 50.0)	9.1 (0.0, 45.5)
	Day 3	Mean ± SdtDev	15.0 ± 13.5	16.4 ± 15.3	15.1 ± 14.0	13.4 ± 11.8	15.2 ± 13.1
		Median (Min, Max)	11.4 (0.0, 70.5)	11.4 (0.0, 70.5)	11.4 (0.0, 68.2)	11.4 (0.0, 47.7)	9.1 (0.0, 54.5)
	Day 4	Mean ± SdtDev	12.9 ± 12.6	14.9 ± 14.2	13.7 ± 14.3	11.2 ± 10.9	12.2 ± 10.7
		Median (Min, Max)	9.1 (0.0, 70.5)	11.4 (0.0, 61.4)	9.1 (0.0, 70.5)	9.1 (0.0, 45.5)	9.1 (0.0, 45.5)
	Day 5	Mean ± SdtDev	9.8 ± 10.6	11.4 ± 12.9	10.5 ± 12.4	8.2 ± 8.5	9.4 ± 8.3
		Median (Min, Max)	6.8 (0.0, 59.1)	6.8 (0.0, 59.1)	6.8 (0.0, 59.1)	5.7 (0.0, 29.5)	8.0 (0.0, 36.4)

Table Key: Group A = Placebo and Psychologic Intervention; Group B = Placebo and General Education; Group C = Propranolol and Psychologic Intervention; Group D = Propranolol and General Education.

On the last day of onsite observation (day 5), participants were asked whether they received propranolol or placebo and no differences were detected in these frequencies (χ2 = 2.72, df = 2, p=0.255). Of those receiving the placebo, 38.2% (n=42) indicated they received the placebo, 30.0% (n=33) indicated propranolol, and 31.8% (n=35) were uncertain. Of those receiving propranolol, 28.8% (n=34) indicated they received the placebo, 30.5% (n=36) indicated propranolol, and 40.7% (n=48) were uncertain.

Primary Outcome

The observed and predicted shoulder pain intensity recovery rates for each intervention group are reported in Table 4. These rates ranged from 55.4% to 62.9% with no statistical differences based on the randomly assigned four intervention groups (p=0.559). In addition, no statistical differences were found in the two group comparisons of those receiving propranolol or placebo (p=0.841) or in the psychologic or general education intervention (p=0.666). The logistic regression results are reported in Table 5 for models that included age, sex, race, and the randomly assigned treatment group. In these models, the placebo and general education treatment group was used as a reference; no differences were found between that group and any of the other randomly assigned treatment conditions (Table 5). The individual treatment group of propranolol and psychologic intervention was hypothesized to be the most efficacious for this high-risk subgroup, but this group’s recovery rate was not better than the reference (OR=0.923, 95% CI=0.427 to 1.996). In addition, no differences were observed in the shoulder pain intensity recovery rates for the combined groups of those receiving propranolol compared with placebo (OR=0.874, 95% CI=0.294 to 2.601) or psychologic intervention compared with general education (OR=1.377, 95% CI=0.466 to 4.068).

Table 4.

Summary of shoulder pain intensity recovery outcomes

Outcome		Overall	Group A	Group B	Group C	Group D	P Value	Placebo (A+B)	Propranolol (C+D)	P Value	Psych Intervention (A+C)	General Education (B+D)	P Value
Days with pain	Median (Min, Max)	4 (0, 11)	4 (0, 11)	4 (1, 11)	4 (4, 9)	4 (4, 9)	0.836	4 (0, 11)	4 (4, 9)	0.841	4 (0, 11)	4 (1, 11)	0.666
	All Participants
Recovered (completed)	No	97 (42.2%)	24 (43.6%)	25 (44.6%)	23 (37.1%)	25 (43.9%)	0.823	49 (44.1%)	48 (40.3%)	0.559	47 (40.2%)	50 (44.2%)	0.531
	Yes	133 (57.8%)	31 (56.4%)	31 (55.4%)	39 (62.9%)	32 (56.1%)		62 (55.9%)	71 (59.7%)		70 (59.8%)	63 (55.8%)
Recovered (Predicted)	No	98 (42.1%)	25 (43.9%)	25 (43.9%)	23 (37.1%)	25 (43.9%)	0.837	50 (43.9%)	48 (40.3%)	0.586	48 (40.3%)	50 (43.9%)	0.586
	Yes	135 (57.9%)	32 (56.1%)	32 (56.1%)	39 (62.9%)	32 (56.1%)		64 (56.1%)	71 (59.7%)		71 (59.7%)	64 (56.1%)
	Participants with PCS Scores ≥ 12
Recovered (completed)	No	54 (46.2%)	13 (46.4%)	15 (46.9%)	12 (40.0%)	14 (51.9%)	0.845	28 (46.7%)	26 (45.6%)	0.909	25 (43.1%)	29 (49.2%)	0.512
	Yes	63 (53.8%)	15 (53.6%)	17 (53.1%)	18 (60.0%)	13 (48.1%)		32 (53.3%)	31 (54.4%)		33 (56.9%)	30 (50.8%)
Recovered (Predicted)	No	55 (46.6%)	14 (48.3%)	15 (46.9%)	12 (40.0%)	14 (51.9%)	0.836	29 (47.5%)	26 (45.6%)	0.834	26 (44.1%)	29 (49.2%)	0.580
	Yes	63 (53.4%)	15 (51.7%)	17 (53.1%)	18 (60.0%)	13 (48.1%)		32 (52.5%)	31 (54.4%)		33 (55.9%)	30 (50.8%)

Table Key: Group A = Placebo and Psychologic Intervention; Group B = Placebo and General Education; Group C = Propranolol and Psychologic Intervention; Group D = Propranolol and General Education.

PCS, Pain Catastrophizing Scale.

Table 5.

Logistic regression models for predicting shoulder pain intensity recovery

DataSet	Variable	Class value	Estimate	StdErr	P value	Odds ratio	OR 95% CI
Complete	Intercept		−0.940	0.991	0.343
	Age		0.052	0.045	0.244	1.054	0.965	1.151
	Sex	Female	−0.560	0.302	0.063	0.571	0.316	1.031
		Male	REF
	Race	Asian	0.674	0.384	0.079	1.962	0.925	4.162
		White	0.655	0.356	0.066	1.925	0.959	3.866
		Black or Other	REF
	Random Group	A	0.012	0.395	0.975	1.012	0.467	2.194
		C	0.227	0.390	0.560	1.255	0.584	2.697
		D	−0.080	0.394	0.838	0.923	0.427	1.996
		B	REF
		A+B vs C+D	−0.135	0.557	0.809	0.874	0.294	2.601
		A+C vs B+D	0.320	0.553	0.563	1.377	0.466	4.068
	Sex*Group	Female*A	−0.386	0.903	0.669
		Female*C	−0.644	0.895	0.472
		Female*D	0.037	0.874	0.966
		Female*B	REF
Predicted	Intercept		−0.0893	0.7787	0.9090
	Age		0.0074	0.0329	0.8220	1.007	0.945	1.075
	Sex	Female	−0.4734	0.2975	0.1120	0.623	0.348	1.116
		Male	REF
	Race	Asian	0.7523	0.3807	0.0480	2.122	1.006	4.475
		White	0.6809	0.3536	0.0540	1.976	0.988	3.950
		Black or Other	REF
	Random Group	A	−0.01617	0.38921	0.967	0.984	0.459	2.110
		C	0.23727	0.38759	0.54	1.268	0.593	2.710
		D	−0.11024	0.39098	0.778	0.896	0.416	1.927
		B	REF
		A+B vs C+D	−0.143	0.550	0.794	0.867	0.295	2.545
		A+C vs B+D	0.331	0.549	0.546	1.393	0.475	4.088
	Sex*Group	Female*A	−0.228	0.886	0.797
		Female*C	−0.654	0.895	0.465
		Female*D	−0.062	0.875	0.943
		Female*B	REF

Table Key: Group A = Placebo and Psychologic Intervention; Group B = Placebo and General Education; Group C = Propranolol and Psychologic Intervention; Group D = Propranolol and General Education.

CI, confidence interval, OR, odds ratio.

Additional Analyses

Our original analysis plan provided flexibility to explore trial findings in the event of null primary findings. These analyses investigated whether there were treatment interactions with sex, self-identified race, and individual level of PCS. None of these interactions contributed statistically to the logistic regression model. Because the sex and treatment interaction was planned given null results, we report that interaction in Table 5.

Table 4 provided descriptive data on the potential impact of PCS: the recovery rates from individuals with higher than the median level of PCS were slightly lower than those from the whole study sample. However, there was no evidence of correlation between baseline PCS and reduction in the worst pain at day 5 (Spearman rho=0.034, p=0.606), and those below and above median PCS were not different in shoulder pain duration (both with median of 4 days, p=0.161 from Wilcoxon rank sum test).

Survival analysis investigated shoulder pain duration (ie, the number of days with shoulder pain) as a more sensitive measure of recovery (Figure 2). In this analysis, no differences were found between any of the four treatment groups. Specifically, those receiving propranolol and the psychologic intervention reported the same number of days with shoulder pain as those receiving placebo and generalized education (HR=1.014, 95% CI=0.686 to 1.500).

Figure 2.

Survival analysis for days to shoulder pain intensity by randomly assigned treatment group

Education 1 = Psychologic Intervention; Education 2 = General Education

Side Effects and Adverse Events

No serious adverse events were reported (ie, no need to seek immediate medical attention, no hospitalization, and no severe allergic reactions). Side effects were closely monitored and captured by self-report at the beginning of each study session, thus reflecting the prior day response to the study drug. The frequency of side effects from day 2 (ie, the study day with the most side effects reported) is reported in Table 6. No differences were observed in the treatment groups for any of the side effects reported. Overall, the number of side effects reported per individual decreased from a day 2 mean of 1.6 (SD=1.7) to a day 5 mean of 0.7 (SD=1.8) (p=0.019).

Table 6.

Side effects reported during the BISP clinical trial

Symptoms		Overall	Treatment Groups					Placebo vs Propranolol
			Group A	Group B	Group C	Group D	P value	Placebo	Propranolol	P value
Lightheadedness	Yes	49 (21.2%)	14 (25.5%)	12 (21.1%)	12 (19.4%)	11 (19.3%)	0.838	26 (23.2%)	23 (19.3%)	0.470
	No	182 (78.8%)	41 (74.5%)	45 (78.9%)	50 (80.6%)	46 (80.7%)		86 (76.8%)	96 (80.7%)
Slowed heartbeat	Yes	9 (3.9%)	3 (5.5%)	3 (5.3%)	2 (3.2%)	1 (1.8%)	0.732	6 (5.4%)	3 (2.5%)	0.322
	No	222 (96.1%)	52 (94.5%)	54 (94.7%)	60 (96.8%)	56 (98.2%)		106 (94.6%)	116 (97.5%)
Weakness	Yes	76 (32.9%)	19 (34.5%)	19 (33.3%)	22 (35.5%)	16 (28.1%)	0.835	38 (33.9%)	38 (31.9%)	0.747
	No	155 (67.1%)	36 (65.5%)	38 (66.7%)	40 (64.5%)	41 (71.9%)		74 (66.1%)	81 (68.1%)
Fatigue	Yes	118 (51.1%)	29 (52.7%)	28 (49.1%)	29 (46.8%)	32 (56.1%)	0.755	57 (50.9%)	61 (51.3%)	0.955
	No	113 (48.9%)	26 (47.3%)	29 (50.9%)	33 (53.2%)	25 (43.9%)		55 (49.1%)	58 (48.7%)
Gastrointestinal upset	Yes	21 (9.1%)	7 (12.7%)	3 (5.3%)	8 (12.9%)	3 (5.3%)	0.276	10 (8.9%)	11 (9.2%)	0.934
	No	210 (90.9%)	48 (87.3%)	54 (94.7%)	54 (87.1%)	54 (94.7%)		102 (91.1%)	108 (90.8%)
Allergic reactions	Yes	1 (0.4%)	0 (0.0%)	0 (0.0%)	0 (0.0%)	1 (1.8%)	0.732	0 (0.0%)	1 (0.8%)	1.000
	No	230 (99.6%)	55 (100.0%)	57 (100.0%)	62 (100.0%)	56 (98.2%)		112 (100.0%)	118 (99.2%)
Dizziness	Yes	21 (9.1%)	5 (9.1%)	7 (12.3%)	3 (4.8%)	6 (10.5%)	0.506	12 (10.7%)	9 (7.6%)	0.405
	No	210 (90.9%)	50 (90.9%)	50 (87.7%)	59 (95.2%)	51 (89.5%)		100 (89.3%)	110 (92.4%)
Nausea/vomiting	Yes	15 (6.5%)	5 (9.1%)	2 (3.5%)	6 (9.7%)	2 (3.5%)	0.373	7 (6.3%)	8 (6.7%)	0.884
	No	216 (93.5%)	50 (90.9%)	55 (96.5%)	56 (90.3%)	55 (96.5%)		105 (93.8%)	111 (93.3%)
Numbness/tingling in the hands	Yes	7 (3.0%)	2 (3.6%)	2 (3.5%)	2 (3.2%)	1 (1.8%)	0.958	4 (3.6%)	3 (2.5%)	0.715
	No	224 (97.0%)	53 (96.4%)	55 (96.5%)	60 (96.8%)	56 (98.2%)		108 (96.4%)	116 (97.5%)
Sleep problems	Yes	33 (14.3%)	6 (10.9%)	9 (15.8%)	7 (11.3%)	11 (19.3%)	0.524	15 (13.4%)	18 (15.1%)	0.707
	No	198 (85.7%)	49 (89.1%)	48 (84.2%)	55 (88.7%)	46 (80.7%)		97 (86.6%)	101 (84.9%)
Depression	Yes	3 (1.3%)	1 (1.8%)	1 (1.8%)	0 (0.0%)	1 (1.8%)	0.709	2 (1.8%)	1 (0.8%)	0.614
	No	227 (98.7%)	54 (98.2%)	56 (98.2%)	62 (100.0%)	55 (98.2%)		110 (98.2%)	117 (99.2%)
Other	Yes	12 (5.7%)	6 (11.8%)	1 (1.8%)	3 (5.5%)	2 (4.1%)	0.197	7 (6.6%)	5 (4.8%)	0.575
	No	198 (94.3%)	45 (88.2%)	54 (98.2%)	52 (94.5%)	47 (95.9%)		99 (93.4%)	99 (95.2%)

Table Key: Group A = Placebo and Psychologic Intervention; Group B = Placebo and General Education; Group C = Propranolol and Psychologic Intervention; Group D = Propranolol and General Education.

DISCUSSION

Very few treatment models for personalized pain management exist[21] and the BISP pre-clinical trial was completed to establish proof-of-principle of the treatment efficacy for this high-risk subgroup.;. When interpreting these results, careful attention should be paid to the model of personalized pain managed used in the BISP trial. That is, we ecruited a high-risk subgroup based on established genetic and psychologic risk factors and then randomly assigned interventions to determine if treatments matched to these risk factors resulted in better outcomes than unmatched treatments. The treatments delivered in this trial were standardized so there was no further personalization beyond having the treatments be a theoretical match for the components of this high-risk subgroup. Our trial results indicated no advantage of receiving the matched treatment (propranolol and psychologic intervention) for the primary outcome of recovery based on shoulder pain intensity. Additional analyses revealed no benefits of the matched treatment based on interactions with sex, race, and level of pain catastrophizing or for a separate outcome of the number of days experiencing shoulder pain. Despite the null findings, these results do add to the existing literature by reporting on a clinical trial specifically designed to test the efficacy of an standardized intervention that was matched to characteristics of this particular high-risk subgroup.[39]

Strengths and Limitations

The primary strength of this trial was the high internal validity afforded by this exercise-induced muscle injury model in a pre-clinical trial. That is, this approach allowed us to standardize the injury causing shoulder pain and ensure high treatment fidelity, close monitoring of outcomes, and an established link with post-operative shoulder pain. Another notable strength of this study is the recruitment of an a priori identified subgroup and the randomization to four different treatment arms, allowing for full exploration of individual or combined treatment effects. Despite being rarely used in investigations of personalized pain management, these design features increased the rigor of this trial.[21]

There are also limitations to this trial that should be considered when interpreting the results. A notable limitation was the inability to recruit a full cohort sample. We enrolled 261 participants, but the original plan was for 440; we were able to reach 59.3% of the projected sample size. The original power estimations were based on a 20% difference in recovery rates, and we observed a 7% difference between the combined personalized treatment arm and the other three comparators. The observed effect was much smaller than what we used in our power estimates; therefore, even with recruiting a full sample, it was highly unlikely that we would have observed any between-group differences.

There were two primary reasons for not meeting the overall recruitment numbers. First, this trial was stopped for several months due to COVID-19 restrictions and even when recruitment was allowed again, it proved difficult to recruit participants for a study that involved 5 days of an onsite commitment given the pandemic-related restrictions.

Shoulder pain induced via exercise may not be the best model for clinical translation due to complexity of shoulder musculature. Therefore, other models of shoulder pain should be studied as they may lead to a different outcome than we observed. Another limitation to consider when interpreting these results is that interventions were matched to characteristics of the high-risk subgroup, but delivered in a standardized manner. This meant the same dosage of propranolol was received and the same information for the psychologic/change oriented intervention was delivered for all participants. Therefore, there may have been inadequate potency of the treatment for some participants, either as under-dosage of propranolol, inadequate information for the psychologic intervention, or both.

A final limitation for this trial was the young age of our participants (mean age=21.0 years, SD=4.3), which was not intended. While approximately 23% of people 18-30 years of age experience chronic pain[36], the young age of our participants means there is limited generalizability to chronic pain populations, which are typically much older in age.

Comparison With Other Clinical Trials

Limited studies report on the efficacy of propranolol for pain management, personalized or otherwise. The Study of Orofacial Pain and PropRANOlol (SOPPRANO) trial tested the efficacy of propranolol in individuals with temporomandibular disorder and had pain intensity as a primary endpoint; thus providing a potential comparison for BISP trial results. Tchivileva et al.[42] reported on 200 participants randomized to receive either extended-release propranolol (same dosage as BISP trial) or a placebo. These participants were followed for 9 weeks, and change in facial pain intensity was the primary outcome. The primary analysis suggested no difference in the mean reduction in facial pain intensity. The lack of differences in the BISP trial across four treatment groups for shoulder pain intensity recovery converges with what was reported for the primary outcome in SOPPRANO.[42] In secondary analyses of the SOPPRANO trial the proportions of those experiencing at least 30% and 50% reductions in facial pain intensity indicated a slight benefit for those receiving propranolol (Number Needed to Treat [NNT]=6.1, p=0.03).[42] In contrast, we found no evidence of benefit of receiving propranolol in any BISP secondary analyses.

Slade et al.[40] then reported on a subset of SOPPRANO participants who were genotyped for the rs4680 single nucleotide polymorphism in COMT. This subset of SOPPRANO participants provides a direct comparison to the BISP trial, since we selectively recruited those with heightened pain sensitivity based on COMT variation. Similar to the primary trial results for SOPPRANO, no differences in facial pain intensity were found for AA vs. GG homozygotes receiving propranolol or placebo. These results also converge with BISP trial findings, although we analyzed COMT single nucleotide polymporphism at rs6269. There would be 75-80% overlap expected in genotyping of the two regions and therefore individuals included in the BISP trial would have at least the same COMT-based pain sensitivity as those included in this sub-analysis for the SOPPRANO trial. Interestingly and counter to what had been expected, a higher proportion of the GG homozygotes experienced at least a 30% reduction in facial pain intensity in the SOPPRANO trial.[40] The BISP trial only recruited those with AA homozygotes, so we could not perform parallel analyses in GG homozygotes.

The BISP clinical trial adds to the existing literature by testing propranolol efficacy for pain at a different anatomical location than in SOPPRANO. Indeed, other differences of note for the BISP trial included that we tested a personalized pain management approach as a primary goal with the enrollment of a high-risk subgroup. In contrast, SOPPRANO investigated the influence of the COMT genotype (independent of pain catastrophizing) in a subgroup analysis.[40] Another key difference in the trials is that the BISP trial included structured informational interventions (psychologic and general education) in addition to propranolol. SOPPRANO only included a pharmaceutical intervention.[42] Despite the differences between the trials, both found no obvious benefit of propranolol for improving pain intensity outcomes in two different pain models (ie, general effect for temporomandibular disorder and personalized effect for exercise-induced shoulder pain).

Design Considerations

Our prior studies tracking pain outcomes following exercise-induced muscle injury highlight that this model does not elicit persistent pain,[6; 9; 16] thus supporting its ethical use for human subjects research. Relatedly, in this trial the context of receiving treatment could have created non-specific effects that induced faster recovery times for all participants. Indeed, the recovery time was shorter for this cohort when compared with our prior cohorts, which did not include the context of receiving treatment.[16; 20; 35] Therefore, it could be that the inherent nature of inducing pain via exercise-induced muscle injury has an upper limit in its suitability to test treatments for differences in pain relief. A final design consideration is that we did not include participants with low-risk phenotype so these results do not speak to the absolute efficacy of the personalized treatments studied.[21]

Future Directions

If efficacy had been demonstrated in this trial, the next logical step would be to plan for a randomized trial in a clinical population. Although that was not the case, there are options for future research. One area for future research would be to investigate the prognostic value of this high-risk subgroup in surgical procedures beyond rotator cuff repair and for other clinically relevant outcomes like sleep interference. The individual factors in the phenotype are not specific to shoulder pain intensity; therefore, they could be predictive of other outcomes for other surgical populations. A high priority would be to consider studying this high-risk phenotype in high-volume orthopaedic procedures with appreciable levels of post-operative pain.[34]For example, a recent study in total joint arthroplasty suggests that approximately 10% of those receiving hip, knee, or shoulder replacement reported high-impact pain following their surgery.[15] Determining if this phenotype predicts high-impact chronic pain following total joint arthroplasty would help to better understand if this risk phenotype is specific to rotator cuff repair or has broader prognostic value. Another avenue would be to investigate this risk phenotype in highly prevalent non-operative conditions, such as low back pain, and determine if there is prognostic value for this phenotype outside of surgical populations.

Conclusion

This BISP pre-clinical trial’s results suggest that matched treatments were not efficacious for improving shoulder pain recovery for a high-risk subgroup when shoulder pain was induced via exercise-induced muscle injury. Accordingly, this phenotype should only be used for prognostic purposes until additional trials are completed in clinical populations.