Effectiveness of training physical therapists in pain neuroscience education for patients with chronic spine pain: a cluster-randomized trial

Abstract

Chronic spinal pain poses complex challenges for health care around the world and is in need of effective interventions. Pain neuroscience education (PNE) is a promising intervention hypothesized to improve pain and disability by changing individuals’ beliefs, perceptions, and expectations about pain. Pain neuroscience education has shown promise in small, controlled trials when implemented in tightly controlled situations. Exploration of promising interventions through more pragmatic methodologies is a crucial but understudied step towards improving outcomes in routine clinical care. The purpose was to examine the impact of pragmatic PNE training on clinical outcomes in patients with chronic spine pain. The cluster-randomized clinical trial took place in 45 outpatient physical therapist (PT) clinics. Participants included 108 physical therapists (45 clinics and 16 clusters) and 319 patients. Clusters of PT clinics were randomly assigned to either receive training in PNE or no intervention and continue with usual care (UC). We found no significant differences between groups for our primary outcome at 12 weeks, Patient-Reported Outcomes Measurement Information System Physical Function computer adaptive test {mean difference = 1.05 (95% confidence interval [CI]: −0.73 to 2.83), P = 0.25}. The PNE group demonstrated significant greater improvements in pain self-efficacy at 12 and 2 weeks compared with no intervention (mean difference = 3.65 [95% CI: 0.00-7.29], P = 0.049 and = 3.08 [95% CI: 0.07 to −6.09], P = 0.045, respectively). However, a similar percentage of participants in both control (41.1%) and treatment (44.4%) groups reported having received the treatment per fidelity question (yes or no to pain discussed as a perceived threat) at 2 weeks. Pragmatic PT PNE training and delivery failed to produce significant functional changes in patients with chronic spinal pain but did produce significant improvement in pain self-efficacy over UC PT.

1. Introduction

Chronic spinal pain is a highly prevalent and costly condition. An estimated 26% of Americans have experienced an episode of low back pain (LBP), and 14% have experienced neck pain (NP) both in the past 3 months.¹⁷ Lifetime prevalence of spinal pain ranges from 54% to 80%, and estimated healthcare costs for those with spinal pain are 57% higher than those without in the United States.^39,40 Although many with acute LBP or NP have a favorable prognosis, those who develop chronic pain experience persistent poor health and place a substantial burden on the healthcare system.^12,57 Despite increasing expenditures for spinal pain, there is evidence that its prevalence and associated disability are increasing.¹⁷ In addition, the current opioid epidemic demands effective nonpharmacologic care for these conditions. The Centers for Disease Control recommends primary care providers consider nonpharmacologic interventions including physical therapy as first-line treatments instead of opioid medications.²⁰ Physical therapists (PTs) provide treatments to reduce pain and promote self-management that should include effective methods to educate patients with chronic spinal pain.¹⁸

Pain research has led to a paradigm shift from a biomedical management model to a biopsychosocial one. The biopsychosocial model more accurately represents how pain is processed by the brain and impacted by contextual, cognitive, emotional, and environmental factor. Although the field of pain science increasingly supports the biopsychosocial model, many patients and providers retain a biomedical perspective focused on tissue damage as directly responsible for, and proportional to, their experience of chronic spinal pain.⁶²

Implementation of educational strategies aimed at improving a patient’s understanding of the biopsychosocial model has been slow. Pain neuroscience education (PNE) is an education method used by clinicians to help patients understand the biology, physiology, and psychological factors that influence the pain experience.^2,32,42–44 Pain neuroscience education is designed to be integrated within a comprehensive plan of care, with an ongoing discussion of topics throughout care instead of a separate educational session.³⁴ Patient education using PNE may aid the patient in reconceptualizing their pain in a biopsychosocial model, which may reduce negative cognitions, promote confidence in self-management, and improve function.⁴²

Pain neuroscience education has been shown to have positive effects on pain intensity,^{36,42,44,46,61} disability,^2,35,42 fear of pain, pain catastrophizing,^2,41–43 physical movement,^{2,35,42,43,47} beliefs about pain,^36,47 beliefs about physical therapy,^41,61 and healthcare costs.³³ These studies are generally focused on efficacy, conducted in controlled settings and using expert clinicians to provide PNE treatment. Although PNE has shown promise in efficacy trials, moving from efficacy to effectiveness is a crucial step toward improving outcomes in routine clinical care. Training clinicians in the use of PNE can improve knowledge and the skills to deliver pain science education. It is unclear if this will translate to changes in patient-centered outcomes. The purpose of this study was to determine the effectiveness of providing PTs with PNE training on patient-centered outcomes for patients with chronic neck or back pain receiving PT under routine clinical circumstances.

2. Methods

2.1. Study design and participants

The study protocol was approved by the University of Utah Institutional Review Board. The trial was registered (Clinicaltrials.gov NCT03168165), and the protocol is published elsewhere.²⁷ The study design was a cluster-randomized clinical trial, randomly assigning geographically and administratively related groups of PT clinics to either receive training in a PNE approach (PNE) or to usual care (UC) with no additional PT training. Participating PT clinics were locations of BenchMark Rehab Partners, which had over 250 outpatient PT clinics in 12 states at the start of the project. For this study, we included only outpatient clinics in the Atlanta, Georgia, and Birmingham, Alabama area, to facilitate onsite training.

Potentially eligible PTs within clinic groups were identified by regional managers then contacted by the study team to inform them about study details and highlight that participation had no impact on employment. Consenting PTs received the training randomly assigned to their clinic group. Outcomes were assessed on consenting patients with chronic NP or LBP treated in these clinics. Patients meeting our eligibility criteria who scheduled an initial PT session at a participating clinic were informed of the study by email then contacted by study staff after a 3-day opt-out period. Patients were recruited from May 2017 through February 2019. Because the intervention (PNE or UC) was provided to PTs, patients were asked to consent to provide outcome measures, which is consistent with a cluster-randomized design. Other than training to provide PNE, there were no attempts to direct PT care in any ways as a result of participation in this study.

For enrolled patients, baseline measures were taken before the first PT session with follow-up assessments at 2 and 12 weeks after baseline. All assessments involved self-report measures collected using Research Electronic Data Capture, a web-based and Health Insurance Portability and Accountability Act-compliant data collection platform.²⁴

2.2. Participants

Eligible patients were between 18 and 75 years of age at the time of their first PT session and primarily seeking care for chronic LBP or NP as per the NIH definition of chronic pain¹⁵ (ie, NP or LBP on at least half the days in the past 6 months). Patients were excluded if they had PT for spinal pain within the previous 6 months, spinal surgery within the previous 12 months, evidence of “red flag” conditions (eg, cauda equine syndrome, cancer, fracture, infection, or systemic disease) that requires immediate referral to medical care, or were knowingly pregnant at the time of recruitment.

Physical therapists’ eligibility criteria included no prior postprofessional training in PNE, no expectation to move or leave the company during the study recruitment period, and not holding a floating position that crossed between PNE and UC regions.

2.3. Randomization

Clinical regions were randomized to either PNE or UC. Randomization was at the region level because randomizing at the individual therapist or clinic level would increase the risk of contamination given the possibility of a patient being cared for by more than one clinician within the same clinic over the course of care and the possibility of therapists working in multiple clinics within a region, respectively. A randomization schedule was developed before enrolment by study statisticians and was stratified by clinic region size, patient volume, and a measure of the region’s clinical outcomes for LBP. Sixteen clinic regions were included with an average of 2.8 clinics per region.

2.4. Interventions

Physical therapist in the PNE assigned clusters received a 16-hour PNE training program. Eight hours of training was provided online, then supplemented with a 1-day, in-person, practice session, designed to improve carryover into clinical practice. Detailed information of the information included in the training is included in our protocol publication.²⁷ Physical therapists were considered successfully trained when they scored >90% on the Neurophysiology of Pain Questionnaire (NPQ), scored >90% on an additional 10-item quiz developed by our training team, and acceptable performance of delivering PNE to mock patients in our live course. Physical therapists in UC-assigned clusters were informed about the study and that their patients may be recruited. No additional training or other interventions were provided to PTs in the UC group. All participating PTs were informed that their patients may be recruited for this study but were not directly informed about which patients were enrolled.

2.5. Outcomes assessments

Baseline data were collected before a patient’s initial PT session. Data were collected at baseline, 2 weeks and 12 weeks remotely by Research Electronic Data Capture. At baseline, patients supplied demographic information including age, sex, race/ethnicity, employment status, general medical information and current history of symptoms, and fear avoidance beliefs measured with the Fear Avoidance Beliefs Questionnaire.⁶²

The primary outcome was the Patient-Reported Outcomes Measurement Information System (PROMIS) Physical Function (PF)-computer adaptive test (CAT) at 12 weeks. The PF-CAT assesses an individual’s capability to perform physical tasks and produces a T score with higher numbers indicating greater function.⁹ The scale that has excellent reliability and validity is highly responsive to change and able to detect a 1.2% difference at 80% power.²¹ The PROMIS PF-CAT was chosen as the primary outcome for several reasons. First, we wanted to include those with chronic spinal pain to include cervical and lumbar pain. Many research studies use separate outcome measures for the different types of spinal pain but doing so makes pooling those results difficult to interpret. The PROMIS PF-CAT shows moderate-to-strong correlation to traditional legacy measures (Oswestry Disability Index and the Neck Disability Index).⁵¹ The CAT form of the PROMIS PF involves fewer items thereby decreasing participant burden.⁵¹ Also, the PROMIS measures have a larger range of measurement than traditional measures, minimizing floor or ceiling effects.²¹ The PROMIS scores are reported T scores that range from 0 to 100, with higher scores indicating higher physical function The general population mean is 50 (SD = 10). Deyo and colleagues used the PROMIS PF-CAT to describe patients with chronic musculoskeletal pain and reported a mean score of 40.9 (SD = 6.7).¹⁶ The authors also described the responsiveness of the PROMIS PF-CAT to aid in the interpretation of change. Over 3 months, patients who reported either “much less pain” and “much worse pain” had +3.85 (SD = 0.85) and −3.85 (SD = −0.72) points of change in the PROMIS PF-CAT, respectively.¹⁶

Secondary outcomes included the PROMIS pain interference CAT,⁹ Numerical Pain Rating Scale assessing pain intensity,^10,11 Pain Self-Efficacy Questionnaire (PSEQ) assessing patients’ confidence in their ability to do daily activities despite pain,⁴⁹ Treatment Self-Regulation Questionnaire to measure the degree of autonomous motivation to follow a treatment regimen and engage in healthy behaviors,^31,56 Pain Catastrophizing Scale assessing negative pain cognitions,^50,58 Working Alliance Theory of Change Inventory to measure therapeutic alliance between a patient and PT,^19,23 and the NPQ to measure knowledge of pain physiology.⁴⁷ In addition, all patients were asked at the 2-week assessment, “Did your therapist discuss your pain with you as an indicator of a perceived threat rather than an indicator of damage or injury to the tissues?” This question was meant to serve as a measure of fidelity that one of the key messages of PNE was delivered to the patient.

2.6. Physical therapist self-report measures

At the time of training, general demographic and practice information were collected, including years of experience, advanced degrees or training, etc. We also collected the NPQ and Health Care Provider’s Pain and Impact Relationship Scale (HC-PAIRS) to assess PT’s attitudes and beliefs about pain. The HC-PAIRS measures healthcare providers’ beliefs about the relationship of a patient’s pain and their physical impairments.^25,29,53,54

2.7. Sample size estimation

We determined our sample size to detect a clinically important effect on the PROMIS PF-CAT after 12 weeks with 80% power while accounting for clustering because of clinic regions. Owing to a lack of published data on meaningful differences for the PF-CAT at the time we developed our protocol, we powered the study to detect a moderate standardized effect (d = 0.40). We assumed an intracluster correlation of 0.025^1,26 to account for clustering within clinic regions and a follow-up rate of 90% for the primary outcome. Under these assumptions, with 16 randomized regions, 319 enrolled patients were required to obtain approximately 80% power with 2-sided α = 0.05 to detect a moderate effect for the PF-CAT between groups.

2.8. Data analysis

Pretreatment baseline patient and PT characteristics were compared between randomized groups to assess chance imbalances in consideration of the cluster-randomized design. In accordance with intention-to-treat principle, all patients were analyzed in the group to which their clinic was randomized regardless of compliance or fidelity. Significance was evaluated using a 2-sided α = 0.05 for both primary and secondary outcomes without adjustment for multiple comparisons.

We began with a series of mixed models to assess the impact of clustering in our data by clinic region and by PT. Differences between groups for the primary outcome (PF-CAT at 12 weeks) were examined using linear mixed model including the baseline PF-CAT score as a fixed-effect covariate. We initially fit mixed models with nested random effects for both PT and clinic region. These models resulted in a variance component estimate of 0 for the clinic region random effect (Appendix Table 1, available as supplemental digital content at http://links.lww.com/PAIN/B458); hence, our final analyses incorporate a single random effect for PT. Models were first performed with adjustment for the baseline level of the outcome only and then with adjustment for the baseline level of the outcome as well as for patient age, sex, race, baseline NPQ scores, baseline worst pain, and baseline overall Pain Catastrophizing Scale scores to account for baseline knowledge of pain and known prognostic variables. Secondary outcomes used similar model to examine between-group differences for the PF-CAT at 2 weeks and for secondary outcomes at 2 and 12 weeks. The same approach was used for all secondary outcomes except the Working Alliance Theory of Change Inventory, which was not assessed at baseline. All but 52 participants had nonmissing values for all baseline covariates. Each analysis was performed on participants with nonmissing values of both baseline and follow-up measurements used in that analysis, without imputation of missing data.

Finally, to explore if the number of visits impacted outcomes at 12 weeks, we compared the number of treatment sessions attended by treatment groups. We tested for an association with the number of treatment sessions by the percent change in PF-CAT (from baseline to 12 weeks) by the treatment group.

All analyses were performed in SAS Version 9.4.

3. Results

Between May 2017 and February 2019, 1884 potentially eligible patients were contacted, of which 443 (23.5%) were eligible and initially agreed to participate (Fig. 1). Ultimately, 319 patients provided consent and began baseline questionnaires. Of these 319 patients, 258 (80.9%) provided the 2-week assessment and 248 (77.7%) provided a 12-week assessment.

Figure 1.

CONSORT flow diagram. PNE, pain neuroscience education; PT, physical therapist; UC, usual care.

One hundred and seventeen PTs were eligible for the study, and 115 consented to participate. The PT demographic and practice history are reported in Table 1. The mean age of PTs was 30.8 (SD = 6.2) with 4.7 (SD = 5.9) years of experience. Baseline characteristics of the patient sample are summarized in Table 2. The mean age of patients was 51.0 (SD = 14.8) years, and the majority were female (67.7%), White (75.1%), and non-Hispanic (95.7%). Most of the samples reported a history of NP or LBP of longer than 6 months (76.8%) and had received treatment previously (62.1%). Patient characteristics were generally similar between groups.

Table 1.

Physical therapist baseline characteristics.

Characteristic	Usual care group (n = 69)	Pain neuroscience education group (n = 56)	P
Age, y (SD)	31.9 (8.0)	30.0 (4.2)	0.163
Sex, n (% male)	20 (29.5)	21 (37.5)	0.407
Years in orthopedic physical therapy practice (SD)	5.7 (6.7)	3.3 (3.6)	0.036
Specialty certifications (% yes)	13.6	17.8	0.567
NPQ (score range: 0-12) (SD)	9.6 (1.1)	9.7 (1.3)	0.803
HC-PAIRS (score range: 15-105) (SD)	50.9 (7.2)	52.2 (7.9)	0.324

HC-PAIRS, Health Care Providers’ Pain and Impact Relationship Scale; NPQ, Neurophysiology of Pain Questionnaire.

Table 2.

Characteristics of the patient sample by the treatment group.

Characteristic	All patients (n = 319)	Usual care group (n = 184)	Pain neuroscience education group (n = 135)
Age, y (SD)	51.0 (14.7)	50.2 (14.3)	52.0 (15.1)
Sex, n (% male)	103 (32.3)	55 (29.9)	48 (35.6)
BMI, kg/m2 (SD)	29.3 (8.2)	30.6 (8.3)	27.7 (7.8)
Obesity, n (% BMI > 30)	118 (39.2)	75 (43.9)	43 (33.1)
Marital status, n (% married or live with a significant other)	220 (69.0)	129 (70.1)	91 (67.4)
Education, n (% with college degree)	177 (55.8)	106 (58.2)	71 (52.6)
Race/ethnicity, n (%)
American Indian/Alaskan Native	3 (0.9)	1 (0.5)	2 (1.5)
Asian	11 (3.5)	8 (4.4)	3 (2.2)
Black/African American	57 (18.0)	41 (22.4)	16 (11.9)
White/Anglo-American	238 (75.1)	129 (70.5)	109 (81.3)
Others	8 (2.5)	4 (2.2)	4 (3.0)
Hispanic or Latino ethnicity	12 (4.3)	8 (5.0)	4 (3.3)
Current smoker, n (%)	15 (4.7)	8 (4.3)	7 (5.2)
History of depression or anxiety, n (%)	127 (40.8)	75 (41.7)	52 (39.7)
Duration of back/NP, mo, N (%)
< 1	8 (2.5)	6 (3.3)	2 (1.5)
1-3	25 (8.0)	17 (9.3)	8 (6.1)
3-6	40 (12.7)	22 (12.1)	18 (13.6)
> 6	241 (76.8)	137 (75.3)	104 (78.8)
Receipt of prior treatment, N (%)
No	120 (37.9)	72 (39.1)	48 (36.1)
Yes	197 (62.1)	112 (60.9)	85 (63.9)
Worst intensity of low back or NP in the past 24 h (0-10 scale)	6.6 (2.2)	6.6 (2.3)	6.6 (2.1)
PCS score	20.0 (12.8)	19.9 (12.9)	20.2 (12.6)
Total NPQ	5.2 (2.4)	5.3 (2.5)	5.2 (2.3)
FABQ, physical activity	14.1 (6.0)	13.6 (5.9)	14.7 (6.0)
FABQ, work	12.2 (11.3)	11.3 (10.9)	13.5 (11.6)
PSEQ	35.3 (14.9)	35.9 (15.4)	34.45 (14.2)

Missing data information: 18 participants missing BMI; 2 missing education; 2 missing race; 37 missing ethnicity; 1 missing smoking; 8 missing depression/anxiety; 5 missing pain duration; 2 missing previous treatment; 15 missing worst level of pain; 18 missing best level of pain; 26 missing PCS; 33 missing NPQ; 30 missing FABQ.

BMI, body mass index; FABQ, Fear Avoidance Belief Questionnaire; NP, neck pain; NPQ, Neurophysiology of Pain Questionnaire; PCS, Pain Catastrophizing Scale; PSEQ, Pain Self-Efficacy Questionnaire.

Physical therapists in the PNE training group improved both their NPQ score (from 9.74 [SD = 1.3] to 11.6 [SD = 0.01], P < 0.01) and HC-PAIR score (from 52.2 [SD = 7.9] to 44.9 [SD = 5.1], P < 0.01) posttraining.

Patients were considered a part of the UC or PNE group based on being scheduled with a participating PT at the time of enrollment. On this basis, 184 of the enrolled patients (57.7%) were in the UC group and 135 (42.3%) were in the PNE group. Forty patients canceled their initial PT session and did not receive treatment (21 [11.4%] from the UC group and 19 [14.1%] from the PNE group). Fifteen patients were scheduled with one PT but saw a different PT for their initial session did not receive the same training. Two patients scheduled with a UC PT actually saw a PNE-trained PT, whereas 13 patients scheduled with a PNE PT saw a UC PT (Fig. 1).

3.1. Primary and secondary outcomes

Results from unadjusted models are shown in Table 3. For the primary outcome, we did not observe differences in PF-CAT by the treatment group at 12 weeks (mean difference = 1.05 [95% confidence interval [CI]: −0.73 to 2.83], P = 0.25) or at 2 weeks (PF-CAT mean difference = 0.29 [95% CI: −0.77 to 1.34], P = 0.59).

Table 3.

Unadjusted mixed models comparing outcome measures between pain neuroscience education and usual care groups at 12 weeks and 2 weeks.

Outcome	Model, wk	Model, n	Mean PNE	Mean UC	Mean group difference	P
PROMIS, PF	12	248	46.50	45.45	1.05 (−0.73 to 2.83)	0.246
	2	258	43.28	42.99	0.29 (−0.77 to 1.34)	0.590
PROMIS, pain	12	191	55.75	57.08	−1.33 (−3.33 to 0.66)	0.188
	2	198	59.12	58.85	0.27 (−1.08 to 1.61)	0.695
PSEQ	12	219	43.33	39.68	3.65 (0.00 to 7.29)	0.0498
	2	236	41.51	38.43	3.08 (0.07 to 6.09)	0.045
TSRQ, autonomous	12	240	4.90	4.85	0.05 (−0.24 to 0.34)	0.731
	2	254	4.84	4.92	−0.08 (−0.38 to 0.22)	0.594
TSRQ, extrinsic	12	240	1.22	1.19	0.02 (−0.27 to 0.31)	0.884
	2	254	1.21	1.17	0.04 (−0.21 to 0.30)	0.744
TSRQ, amotivation	12	239	1.34	1.39	−0.05 (−0.40 to 0.30)	0.795
	2	253	1.31	1.22	0.09 (−0.20 to 0.38)	0.539
WATOCI	12	235	37.92	37.94	−0.02 (−2.13 to 2.09)	0.985
	2	248	39.10	38.80	0.30 (−1.51 to 2.11)	0.742

Models include baseline response (except WATOCI).

PNE, pain neuroscience education; PROMIS, Patient-Reported Outcomes Measurement Information System; PSEQ, Pain Self-Efficacy Questionnaire; TRSQ, Treatment Self-Regulation Questionnaire; UC, usual care; WATOCI, Working Alliance Theory of Change Inventory.

Evaluation of secondary outcome measures did not find any between-group differences with the exception of the PSEQ. At both 12 and 2 weeks, patients in the PNE group reported higher PSEQ scores than patients in the UC group (mean difference = 3.65 [95% CI: 0.00-7.29], P = 0.049 and = 3.08 [95% CI: 0.07 to −6.09], P = 0.045, respectively). Results of fully adjusted models were comparable with the unadjusted results (Appendix Table 2, available as supplemental digital content at http://links.lww.com/PAIN/B458).

3.2. Fidelity measure and physical therapy visits

At 2 weeks, 108 of 254 overall patients (42.5%) responded “yes” to the fidelity measure. In the PNE group, 48 of 108 patients (44.4%) responded “yes,” whereas in the UC group, 60 of 146 patients (41.1%) responded “yes.”

We also used explored the potential influence of the number of visits received on our main outcome between groups. The mean number of physical therapy visits attended by enrolled patients was 10.0 visits (SD = 6.9, range 0-39). Patients in the UC group had a mean = 9.0 (SD = 6.0, range 1-37) visits. Patients in the PNE group had a mean = 11.4 (SD = 7.8, range 0-39) (mean difference 2.4,95% CI: 0.8-4.1, P = 0.004) visits. To explore if the number of visits may have impacted change in PF-CAT scores, the number of treatment sessions by the percent change in PF-CAT (from baseline to 12 weeks) is graphed in Appendix Figure 1 (available as supplemental digital content at http://links.lww.com/PAIN/B458). In both treatment groups, we did not observe an association of the number of treatment sessions with changes in PF-CAT scores (P = 0.94 for PNE, P = 0.79 for UC) (Appendix Table 3, available as supplemental digital content at http://links.lww.com/PAIN/B458).

4. Discussion

This cluster-randomized clinical trial examined the effectiveness of training PTs to provide PNE for patients with chronic neck or back pain. We enrolled 115 PTs across 16 clinic regions and 319 patients who were scheduled to receive PT. The primary outcome, physical function measured by PROMIS PF-CAT at 12 weeks, failed to demonstrate a difference between the PNE and UC groups. Secondary outcome measures were also not significantly different at 2 or 12 weeks with the exception of pain self-efficacy, which favored patients treated by a PT trained in PNE. There are a few possible explanations for the lack of difference between groups: (1) our training did not result in a significant behavior change for therapists, (2) a number of patients never received therapy or cross-over between groups, or (3) the intervention was not effective for this group of patients.

Pain neuroscience education is recommended as a part of comprehensive treatment for those with LBP and NP^5,6 and has shown promise in controlled efficacy trials.³⁷ Our study used a pragmatic approach and failed to find differences favoring PNE training for most outcomes. Several other studies have also failed to find a statistical or clinically important difference in function using pain education provided using a variety of strategies.^{4,22,33,38,45,48,52,55,59,60} Collectively, these findings highlight the well-known challenges in translating efficacious treatments into real-world clinical care.

It is not certain if the training provided resulted in a behavior change for therapists. The therapists who participated in the training did demonstrate a statistically significant improvement in their knowledge of pain and their beliefs and attitudes about pain, using the NPQ and HC-PAIRS, respectively. Clinically meaningful change values for the NPQ or HC-PAIRS are not established. Although we did exclude PTs who had completed formal PNE training, the baseline NPQ and HC-PAIRS scores for both PNE and UC groups were fairly high. We did not make an effort to eliminate those who may have completed some form of self-study into PNE. The group of PTs enrolled was young, having recently completing their professional training and demonstrated high baseline knowledge of pain from the biopsychosocial perspective as evidenced by the NPQ scores. Because of PNE’s popularity in the PT profession, it is possible that information on the biopsychosocial nature of pain is increasingly integrated into entry-level education, other continuing education, or gained through books, videos, and other materials in the form of self-study. For example, a recent study of PT students found very high pain knowledge was present during the professional education process.¹³ There was no formal integration of PNE from a company or regionlevel, but some clinics may have informally developed a form of PNE for their patients. At the beginning of this trial, it was a common belief that not many providers had adopted an approach consistent with PNE; however, this may have changed over the course of the study.

Another potential issue may have resulted from the high levels of pain knowledge assessed with the NPQ. Although a previous study reported the NPQ demonstrates no ceiling effect,⁸ we found 19% of PTs were within 1 point of the maximal score. This may reflect that the PTs in this study already had a very good knowledge of pain. It is also possible our training dosage may have been insufficient to improve knowledge enough to make an appreciable difference in their practice.

Our results may also reflect other common challenges in implementing new practices into routine clinical care. Constraints on the amount of time PTs had to spend with patients may have limited the use of PNE during care. In addition, although the PTs in our study volunteered for participation, they did not self-select PNE as a priority for their continuing education, which may have caused them to perceive the training as a lower value compared with if they selected the training themselves. Participating PTs may not have been as motivated to implement PNE or in the appropriate stage of change to make a meaningful and lasting change in their practice without additional or ongoing support such as additional training, peer-support, performance feedback, mentoring, etc.

Our pragmatic study design made fidelity monitoring challenging. Ideally, patient sessions would have been observed or videotaped to monitor fidelity. This method would have been difficult to integrate into routine care and would not provide information about the patient’s understanding of PNE messages. We chose to inquire if the patient recalled being told a key PNE message as a fidelity measure. Using our measure, we found no difference between groups in the proportion of patients self-reporting that they received a foundational message of PNE. It may be that patients’ recall was inaccurate or that PTs trained in PNE were not effective in consistently conveying key PNE messages.

In our study, several patients never received therapy or crossed-over between groups. However, there were only 6 individuals with PROMIS scores at 12 weeks who received care under the incorrect therapist.

Another reason for the study’s findings may be that PNE treatment was not targeted to the subgroup of patients most likely to benefit. Pain neuroscience education has been used with many different populations and theoretically best targeted at those with chronic pain or to prevent an acute transitioning to chronic pain. It may be that PNE is best targeted to patients with chronic pain who also demonstrate unhelpful or maladaptive pain.

The one outcome difference between groups was pain self-efficacy at short-term and long-term follow-up. Self-efficacy relates to the degree an individual feels they have control over their situation.³ Pain self-efficacy is defined as a patient’s confidence in his or her ability to tolerate pain and perform daily activities despite pain.⁴⁹ Higher self-efficacy has been related to improved outcomes and a propensity for more active coping strategies.^{3,14,28,63,65} Higher self-efficacy is also associated with fewer pain behaviors and lower disability.^7,64 Pain self-efficacy seems to be an important construct for optimizing PT outcomes, particularly for developing self-management capabilities for patients with spinal pain. One other pilot study examining the effectiveness of PNE added to group exercise also failed to find a difference in function but did find difference in self-efficacy.⁵⁵ Self-efficacy has been shown to mediate the relationship between pain and disability, and the ability of PNE to enhance self-efficacy may suggest that more effective PNE delivery might lead to differences in physical function,³⁰ but this was not shown in our study.

There are several limitations to our study. First, our fidelity measure may have been insufficient to assess whether PNE was delivered to the PNE group and not to the UC group. A more robust fidelity assessment for the PNE intervention may have been achieved by using a checklist of key PNE items. Our follow-up period was only 12 weeks. Owing to the nature of chronic pain, longer-term outcomes are preferable. Lengthening this follow-up period may also have elucidated the influence of self-efficacy on coping with the chronic, recurrent nature of neck and back pain. Our PNE training dose may have been insufficient to cause a change in patient-centered outcomes. Finally, we did not evaluate different implementation strategies for PNE in our study.

In conclusion, our pragmatic, cluster-randomized clinical trial on the effectiveness of PNE delivered to those with chronic neck or back pain failed to demonstrate differences in pain or disability, although it did result in a significant improvement of pain self-efficacy. It is likely that PTs were more knowledgeable about the key messages of PNE than in previous inquiries, as evidenced by high baseline pain knowledge. It is also possible that PNE messages are being delivered to patients without formal, postprofessional training in PNE.

Future studies should investigate the implementation of PNE after training, overcoming specific barriers to incorporating PNE into patient care for PTs, and identify the characteristics of strong adopters.