Restorative Neurostimulation Therapy Compared to Optimal Medical Management: A Randomized Evaluation (RESTORE) for the Treatment of Chronic Mechanical Low Back Pain due to Multifidus Dysfunction

Abstract

Introduction

Many interventional strategies are commonly used to treat chronic low back pain (CLBP), though few are specifically intended to target the distinct underlying pathomechanisms causing low back pain. Restorative neurostimulation has been suggested as a specific treatment for mechanical CLBP resulting from multifidus dysfunction. In this randomized controlled trial, we report outcomes from a cohort of patients with CLBP associated with multifidus dysfunction treated with restorative neurostimulation compared to those randomized to a control group receiving optimal medical management (OMM) over 1 year.

Methods

RESTORE is a multicenter, open-label randomized controlled trial. Candidates were assessed for CLBP associated with multifidus dysfunction, with no indication for or history of lumbar spine surgery. Participants were randomized to either restorative neurostimulation with the ReActiv8 system or OMM. The primary endpoint was a comparison of the mean change in the Oswestry Disability Index (ODI) between the treatment and control arms at 1 year, and secondary endpoints included pain (numeric rating scale [NRS]) and health-related quality of life (EuroQol Five-Dimension [EQ-5D-5L]).

Results

A total of 203 patients, average age 47 years, and with an average 11-year history of low back pain, were included in the analysis. The primary endpoint was a statistically significant demonstration of a clinically relevant mean improvement in the Oswestry Disability Index (ODI) between restorative neurostimulation and OMM arms: ODI (−19.7 ± 1.4 vs. −2.9 ± 1.4; p < 0.001). Additionally, improvements in both the numeric rating scale (NRS) (−3.6 ± 0.2 vs. −0.6 ± 0.2; p < 0.001) and EuroQol Five-Dimension (EQ-5D-5L) (0.155 ± 0.012 vs. 0.008 ± 0.012; p < 0.001) were statistically and clinically significant in the restorative neurostimulation arm compared to the OMM arm.

Conclusion

The RESTORE trial demonstrates that restorative neurostimulation is a safe, reversible, clinically effective, and highly durable option for patients suffering with nonoperative CLBP associated with multifidus dysfunction. This demonstration of treatment superiority over OMM through 1 year is a significant milestone in addressing a major health burden and unmet clinical need.

Trial Registration

ClinicalTrials.gov Identifier: NCT04803214.

Plain Language Summary

Chronic low back pain can occur as a consequence of dysfunction in the key stabilizing muscles of the spine, the multifidi. This type of low back pain is difficult to treat, with many interventions resulting in limited improvement or short-term relief for a significant proportion of patients. Despite this limitation, these approaches still represent the best available care in most practices. Restorative neurostimulation is a technique that stimulates dysfunctional multifidi, overriding muscle inhibition to improve spinal function, reduce disability, and alleviate pain. The hypothesis was that this treatment is appropriate for a specific subset of patients who have failed to respond to best available conservative and interventional care. The goal of this study was to compare the effect of restorative neurostimulation to standard-of-care interventions (optimal medical management) for patients with chronic mechanical low back pain. Patients with an average 11-year history of chronic low back pain and diagnosed with multifidus dysfunction were randomly assigned to either ongoing optimal medical management or restorative neurostimulation. At 1 year, disability, pain, and healthcare-related quality of life were assessed. Patients treated with restorative neurostimulation demonstrated significant improvements in their clinical outcomes compared to those receiving optimal medical management alone. Device-related adverse events were rare, reinforcing the safety profile of this technique. This study demonstrated that without restorative neurostimulation, patients with chronic low back pain and multifidus dysfunction have very few effective options and obtained little clinical benefit from ongoing optimal medical management. Restorative neurostimulation is an important advancement for this difficult-to-treat population.

Key Summary Points

Why carry out this study?

This randomized controlled trial was conducted to provide a level 1 comparison between restorative neurostimulation using the ReActiv8 system and optimal medical management (OMM).

What was learned from this study?

Restorative neurostimulation was superior to OMM across all primary, secondary, and tertiary endpoints.

Patients receiving restorative neurostimulation showed significant improvements in disability,pain, and healthcare-related-related quality of life after 1 year of treatment.

The majority of patients treated with OMM derived no benefit or had worsening disability, pain,and health-related quality of life after 1 year.

Restorative neurostimulation should be considered earlier in the care continuum for mechanical chronic low back pain.

The continuation of conventional pain mitigation strategies that fail to yield durable improvements should be reconsidered.

Introduction

The treatment paradigm for chronic mechanical low back pain (CLBP) remains a fragmented application of various medical and interventional approaches often failing in terms of effectiveness and durability [1]. As a result, CLBP has become the most common pain condition in US adults and a leading cause of years lived with disability globally [2]. Little progress has been made in addressing this major component of the global burden of disease [3].

While acute episodes of low back pain are for the most part transient, the initial episode is highly correlated with repeated occurrences of back pain, marked by increased severity and reduced periods of remission [4]. As the low back pain symptoms become chronic, patients with physical disabilities may exhibit maladaptive responses to their pain. In these cases, the impact on a patient’s quality of life often transcends the physical and functional, leading to psychosocial manifestations [5]. For this reason, current best-management strategies should be multimodal and optimized to each patient’s individual needs [6].

Clinical management guidelines vary, but generally recommend beginning with a stratified approach emphasizing conservative care including education, yoga, bracing, massage, and nontraditional medicine (e.g., acupuncture) [7, 8]. Physical and exercise therapies, maximizing the use of non-opioid medications, and education/counseling to address the psychosocial impact of chronic pain using a multidisciplinary approach have shown some benefit. However, such treatments have demonstrated only small to moderate effects on pain and disability, and while they are beneficial in the early stages of low back pain, there tends to be a significant proportion of patients who have transitioned to a chronic pain state that derive little or no improvement over the long term [9]. Consequently, patients and physicians may choose to escalate the invasiveness of their care. Invasive interventional therapies such as facet joint or nerve blocks, epidural steroid injections, and radio frequency ablation often provide an inadequate or only short-term benefit at best [10–14]. Even when more successful, these procedures often have diminishing returns and are rarely associated with significant sustained functional improvements. Repeated interventional procedures represent a huge healthcare cost with minimal long-term benefits [15]. Finally, despite the paucity of evidence supporting the use of traditional spine surgery for non-radicular back pain in the absence of surgical indications such as spinal deformity, stenosis, or instability, many of these patients eventually opt for surgery (e.g., decompression, fusion, or arthroplasty), as few viable alternatives remain [16]. Unfortunately, when mechanical low back pain without other clearly defined pathology is the chief complaint and the primary reason for surgery, the outcomes are generally inconsistent and poor [17].

Treatment options thus remain limited for patients with CLBP who have not responded to this management pathway. This may be due, in part, to the categorization of mechanical or nociceptive pain as “non-specific” low back pain. The use of this “non-specific” term implies an idiopathic etiology, and with a poor understanding of the underlying cause comes poor coordination of healthcare resources directed towards this issue [18, 19]. For example, despite being recognized in the physical therapy literature as a specific phenotype of low back pain, CLBP associated with multifidus dysfunction and lumbar motor control deficit has until recently received little attention [20]. Restorative neurostimulation has been suggested as a specific therapy aimed at treating this particular subset of CLBP. The recent publication of 5-year clinical outcomes from the ReActiv8-B clinical trial demonstrated that such diagnostic specificity in the application of CLBP interventions can result in both durable and clinically meaningful improvements [21]. The ReActiv8-B trial clearly established the efficacy, safety, and durability of restorative neurostimulation for the treatment of mechanical CLBP due to multifidus dysfunction, with durable outcomes including clinically meaningful improvements in pain, function, and health-related quality of life [21].

Despite the published data on the long-term effectiveness of restorative neurostimulation in trials and real-world settings [22–24], however, no study has directly compared it to commonly used chronic low back pain management strategies. The aim of this randomized controlled trial was to compare clinical outcomes in a specific cohort of patients with CLBP associated with multifidus dysfunction treated with restorative neurostimulation to the outcomes from a control arm receiving optimal medical management. Here we report the 1-year outcomes.

Methods

Trial Design

The protocol for the RESTORE clinical trial (ClinicalTrials.gov NCT04803214) was prospectively published according to SPIRIT (Standard Protocol Items: Recommendations for Interventional Trials) guidelines [25]. In brief, RESTORE is a post-market, multicenter, open-label, randomized controlled trial (RCT) performed at 25 clinical sites in the United States, and no changes were made from the published protocol. Patients with moderate to severe pain and disability associated with CLBP and multifidus dysfunction were recruited from the investigator’s clinical practice between July 2021 and July 2023. They were screened against previously published inclusion (Table 1) and exclusion (Table 2) criteria and enrolled. Eligibility was verified by a three-member panel of independent medical monitors prior to randomization. Patients were randomized to either restorative neurostimulation with the ReActiv8 system (treatment) or optimal medical management (OMM) as defined below (control). After the primary endpoint visit at 1 year, patients in the control arm may elect to receive the ReActiv8 system (Mainstay Medical, Dublin, Ireland). All patients are then followed for an additional year. The focus of this paper is on the primary and secondary endpoints at 1 year after randomization. The conduct of the trial complied with the Declaration of Helsinki of 1964 and its later amendments, good clinical practice, and institutional review board (IRB) requirements. The WCG IRB acted as the central institutional review board (IRB) (RN #20211219) for most sites, while other sites had local IRB approval prior to enrollment of patients. All patients provided written informed consent to participate in the trial. IRB approval was obtained at each site, and the results are reported in accordance with the CONSORT (Consolidated Standards of Reporting Trials) guidelines. The trial schema was published previously [25].

Table 1.

Inclusion criteria

1. Age 21 years or older at time of enrollment

2. Evidence of lumbar multifidus muscle dysfunction (both radiological and clinical tests)

3. Intractable chronic low back pain that has persisted longer than 6 months prior to enrollment, resulting in pain most days in the 12 months prior to enrollment

4. Failed therapy including pain medications and physical therapy

5. Not a candidate for spinal surgery

6. Low back pain rated on a numeric rating scale (NRS) of ≥ 6 and ≤ 9

7. Oswestry Disability Index (ODI) score ≥ 30 and ≤ 60

8. Willing and able to provide informed consent

9. Able to comply with study protocol

10. On optimal medical management (per investigator)

Table 2.

Exclusion criteria

1. Contraindicated for the ReActiv8 system

a. Unable to operate the ReActiv8 system

b. Unsuitable for ReActiv8 implant surgery

2. BMI > 35

3. Back pain characteristics:

a. Any surgical correction procedure for scoliosis at any time, or a current clinical diagnosis of moderate to severe scoliosis (Cobb angle ≥ 25°)

b. An independent MRI assessment identifying a pathology that is likely the cause of the chronic low back pain and is amenable to spine surgery

4. Leg pain described as being worse than back pain, or radiculopathy (neuropathic pain) below the knee

5. Surgical or other procedures exclusions:

a. Any previous back surgery (e.g., laminectomy, discectomy, spinal fusion) at or below segmental level T8

b. Any previous thoracic or lumbar sympathectomy

c. Any lumbar medial branch nerve rhizotomies within the past 12 months

d. Any lumbar medial branch nerve blocks within the past 30 days

e. Any previous or existing neuromodulation devices (e.g., drug pump, spinal cord stimulation, and/or peripheral nerve stimulation)

6. Other clinical conditions:

a. Pregnant or planning to be pregnant in the next 12 months

b. Any condition unrelated to chronic low back pain such as muscle wasting, muscle atrophy, or progressive neurological disease which, in the opinion of the investigator, could limit physical movement or compliance with the protocol, or interfere with the assessment of efficacy

c. Evidence of an active disruptive psychological or psychiatric disorder or other known condition significant enough to impact perception of pain, compliance with intervention, and/or ability to evaluate treatment outcome (e.g., active depression, bipolar disease, Alzheimer’s disease) as determined by the investigator in consultation with a psychologist or psychiatrist, as appropriate

d. An opioid addiction or drug-seeking behavior, as determined by the investigator

e. Any active malignant disease

f. Any active infection in the vicinity of the implant site or any systemic infection

g. Poorly controlled diabetes (type I or type II) determined by HbA1c > 8

7. General exclusions:

a. Current smoker

b. Current or planned participation in any other clinical trial during the study

c. A condition currently requiring or likely to require use of MRI or diathermy

d. Life expectancy less than 1 year

e. A pending or approved financial compensation claim (e.g., worker’s compensation claim, long-term disability claim, injury claim under litigation)

BMI body mass index, HbA1c glycated hemoglobin, MRI magnetic resonance imaging

Randomization

Patients were randomized to continuing OMM (control arm) or continuing OMM with ReActiv8 (treatment arm) in a ratio of 1:1 at enrollment. Randomization was performed according to a random permuted block design stratified by clinical site. For concealment, treatment assignment was available through the clinical database and was performed by the study site personnel once the patient had met the enrollment criteria and was approved by the medical monitors to proceed for randomization. The allocation sequence was generated by an independent statistician who was not involved in the conduct of the study.

Sample Size

A minimum of 204 evaluable patients was estimated to sufficiently power the primary endpoint. Power calculations for this trial relied on the following assumptions, based on a prior trial of the ReActiv8 system: minimum power of 80%, type I error rate of 5%, assumed mean change in treatment group of 18.2, assumed mean change in control group of 12.2, and pooled standard deviation of 15.

Assessment of Multifidus Dysfunction

Patients were required to be diagnosed with multifidus muscle dysfunction prior to randomization in the trial. There were four recommended methods, and at least one positive finding was mandated, though discretion was left to the treating physician to determine the most appropriate test given the patient presentation and capacity to undertake physical exams and whether additional tests were required to confirm dysfunction.

Prone Instability Test

The prone instability test (PIT) [26, 27] is conducted with the patient lying face down in a neutral spine position. The examiner applies posterior-to-anterior pressure to each lumbar segment. If one or more segments elicit pain, the test is repeated with the patient engaging their posterior spinal muscles by lifting their feet off the ground, activating hip extension. A positive test, indicated by a significant reduction in pain during muscle activation, suggests a motor control deficit involving multifidus muscle dysfunction.

The Multifidus Lift Test

During the multifidus lift test (MLT) [28, 29], the patient lies prone on an examination table with their elbows aligned with their ears. Lordosis may be minimized using a pillow placed under the abdomen. The examiner locates the multifidus muscle in the posterolateral gutter between the spinous process and the longissimus muscle. The patient is then instructed to lift their arm approximately 10 cm off the table contralateral to where the examiner is palpating. This allows the examiner to directly assess muscle activation, noting any weakness or compensatory activation of other muscles, such as the longissimus.

Aberrant Movements

Assessing the quality of lumbar motion is a critical component of a spinal physical examination, providing insight into motor control deficits and multifidus dysfunction [30, 31]. Five specific movement patterns are particularly important: altered lumbopelvic rhythm, Gower’s sign, sagittal plane deviation, instability catch, and a painful arc of motion. These patterns can help identify underlying impairments in motor coordination and muscle function, including those associated with spinal instability.

Magnetic Resonance Imaging (MRI) of Fat Infiltration

Multiple studies have demonstrated the correlation between the reduction in lean muscle in the multifidus and chronic low back pain, and the observation of muscle degeneration on MRI is a relevant sign of the consequences of muscle inhibition and indicative of a deficit in the motor control strategies available to a patient. Kjaer et al. [32] provide a robust demonstration of the relationship between MRI findings of fat infiltration of the multifidus muscle, specifically grade 1 (10–50%) and grade 2 (> 50%), and CLBP.

A combination of physical tests, MRI findings of fatty infiltration, and a clinical presentation consistent with a mechanical origin of the pain has been shown to be clinically predictive of response to restorative neurostimulation.

Optimal Medical Management

OMM was defined in the protocol as a treatment plan that is individualized to each patient’s specific needs. To be considered part of the treatment plan, the therapy should be in use or have previously been applied. The OMM plan was documented in a standardized format prior to randomization and only considered the use of non-investigational interventions, including pharmacological agents, non-pharmacological, physical, or psychosocial therapies (e.g., physiotherapy, spinal manipulation, exercise programs, and cognitive behavioral therapy), and interventions such as epidural steroid injections, facet nerve blocks, and nerve ablations. If the enrolling physician identified a potentially relevant therapy that had not yet been applied, the patient was not included in the trial until the effect of the therapy had been observed. At this point, patients were considered “optimized” and were able to be enrolled and randomized provided other eligibility criteria were met. Interventions deemed clinically appropriate by the treating physician and included in the OMM plan could be utilized by patients in either arm of the trial at any time.

Control Arm: OMM

Patients randomized to the control arm continued with the OMM treatment plan that was established prior to randomization.

Treatment Arm: Restorative Neurostimulation

Patients randomized to the treatment arm were additionally treated with restorative neurostimulation (ReActiv8). The rationale and technique for implantation have been described elsewhere [22], but briefly, it consists of episodic bilateral motor stimulation of the L2 medial branch of the dorsal ramus as it crosses the L3 transverse process by two distally fixated leads attached to an implantable pulse generator, programmed to a stimulation frequency of 20 Hz, a pulse width of 214 ms, and participant-specific pulse amplitudes and electrode configurations to elicit smooth tetanic contractions of the multifidi. At follow-up visits, stimulation parameters could be adjusted to maintain comfortable contractions, if required. Therapy was initiated by patients with a daily maximum of 60 min of stimulation delivered over a minimum of two sessions per day. This repeated electrical activation of the deep lumbar multifidus overrides underlying neurological inhibition and has been shown to induce structural changes in the muscle consistent with improved function [33]. Patients were instructed to use the device for the full 60 min a day through the 1-year primary endpoint assessment visit.

Follow-Up

Patients in both arms returned at regular intervals for data collection and treatments as needed and specified in the OMM plan. At these visits, patients in the treatment arm may have also received adjustments to their stimulation parameters to ensure that muscle contractions remained comfortable.

Outcomes

The primary endpoint was a comparison of the mean change in the Oswestry Disability Index (ODI) between the treatment and control arms at the 1-year follow-up visit.

Prespecified secondary analyses were a comparison of change from baseline in average low back pain in the last 24 h measured using the 11-point numeric rating scale (NRS), and a comparison of change from baseline in quality of life measured using the EuroQol Five-Dimension Scale (EQ-5D-5L).

Prespecified secondary and tertiary analysis of the completers (i.e., patients with data at the visit) were a comparison of the following measures: mean change in ODI; mean change in NRS; mean change in EQ-5D-5L; percent pain relief, subject global impression of change (SGIC); treatment satisfaction; proportion of patients with a > 15-point ODI improvement and/or a > 50% low back pain NRS improvement and no worsening in either measure; and longitudinal change in medications. Outcome measures were used with permission.

Statistical Analysis

The primary endpoint was analyzed using a mixed model for repeated measures (MMRM), which provides unbiased estimates of treatment effect in the presence of missing data. The model requires at least one follow-up visit for a patient to be included in the analysis. This is performed as a modified intent-to-treat analysis, and the primary endpoint was therefore predefined to consist of randomized patients who completed at least one follow-up visit. The statistical analysis plan required patients lost to follow-up for reasons related to lack of efficacy (LOE) or subsequent to a device/procedure-related adverse event to utilize a baseline-observation-carried-forward estimation of missing values; otherwise, when there was missing follow-up and the missing-at-random (MAR) assumption was true, the missing values were to be implicitly imputed by an MMRM [34]. This model included baseline values (ODI, NRS, or EQ-5D-5L), treatment arm, visit, and visit-by-treatment arm interaction, and changes from baseline to months 1.5, 3, and 6, as well as from baseline to 1 year as a function of the fixed effects. The mixed models were adjusted for the baseline value and included covariates for treatment group, visit, and the interaction between treatment and visit. The reason, assumption, and method of imputation for the missing data are reported in Table 3. Throughout, these results are referred to as “control MMRM” or “treatment MMRM.”

Table 3.

Assumptions for missing data at the primary endpoint

Withdrawal/lost to follow-up reason	Assumption	Arm	N
Elected to pursue ReActiv8 commercially	Not random	Control	1
Explant for reasons other than LOE a	MAR	Treatment	2
Voluntary withdrawal from control arm	MAR	Control	2
Device-related adverse event b	Not random	Treatment	1
Lost to follow-up	MAR	Control	4
Lost to follow-up	MAR	Treatment	2

LOE lack of efficacy, MAR missing at random, subjects “missing not at random” had baseline observation carried forward imputation for missing data; ^aone noncompliance (mental health) and one voluntary withdrawal; ^b infection

Endpoints were also assessed utilizing available data only (completers) at each visit and are presented in addition to the MMRM analysis. Completer cohorts are referred to as control and treatment arms.

Differences between the treatment and control arms are presented with corresponding 95% confidence intervals (CIs). A p-value < 0.05 was considered statistically significant. All statistical analyses were prespecified in a detailed statistical analysis plan and conducted using SAS version 9.3 or later (SAS Institute Inc., Cary, NC, USA) by independent biostatisticians.

Results

Participant Disposition

Of the 386 patients screened for inclusion in this trial, 160 were excluded prior to randomization, and an additional 23 patients withdrew prior to any follow-up visits, resulting in 203 patients in the analysis set: 99 in the treatment arm and 104 in the control arm. At 1 year, 94/99 (94.9%) of the treatment-arm patients and 94/104 (90.3%) of the control arm were available for assessment of outcomes. The full disposition and reasons for missing visits and withdrawal from the trial are reported in the CONSORT diagram in Fig. 1.

Fig. 1.

Flow diagram for participant disposition

Study Population

The baseline demographics (Table 4) between the treatment and control arms were well-balanced, apart from a non-clinically relevant but statistically significant difference in Depression Anxiety and Stress Scale (DASS) [35] and active depression. Overall, patients were on average 47 ± 12 years of age, 62% were female with an average body mass index (BMI) of 29 ± 4 kg m², and were an average of 11.4 ± 8.7 years from the onset of their low back pain. The average ODI was 44 ± 8, back NRS 7.1 ± 0.9, and EQ-5D-5L index 0.619 ± 0.120. All patients had at least one objective finding of multifidus muscle dysfunction by either physical assessment or MRI imaging [36].

Table 4.

Key baseline demographics

Characteristic	Treatment N = 99	Control N = 104	p-value
	Mean ± SD (Min, Max) or Pt (% Pt)	Mean ± SD (Min, Max) or Pt (% Pt)
Age (years)	45 ± 11(21, 68)	48 ± 13 (23, 74)	0.176
Sex
Female	58 (59%)	68 (65%)	0.318
Male	41 (41%)	36 (35%)
Body mass index (BMI)	29 ± 4 (19, 35)	28 ± 4 (19, 35)	0.359
Pain duration (years from first occurrence)	11.8 ± 9.7 (1.2, 39.8)	11.1 ± 7.5 (0.9, 29.3)	0.680
Days with LBP (%)a	95 ± 13 (33, 100)	97 ± 8 (60, 100)	0.128
Leg pain	31 (31%)	28 (27%)	0.491
History of depression	39 (39%)	52 (50%)	0.129
Active depressionb	15 (38%)	32 (62%)	0.029
ODI	44 ± 8 (30, 60)	44 ± 8 (30, 60)	0.966
LBP NRS	7.1 ± 0.8 (6.0, 9.0)	7.1 ± 0.9 (6.0, 9.0)	0.912
EQ-5D-5L index score	0.626 ± 0.116 (0.306, 0.818)	0.612 ± 0.125 (0.315, 0.827)	0.418
DASS c	5.3 ± 5.6 (0.0, 24.0)	7.1 ± 6.2 (0.0, 25.3)	0.040
Evidence of lumbar multifidus dysfunction
At least one positive finding d	99 (100%)	104 (100%)

DASS Depression Anxiety and Stress Scale, EQ-5D-5L EuroQol Five-Dimension Scale, LBP low back pain, NRS numeric rating scale, ODI Oswestry Disability Index, SD standard deviation

^aDays in the past year

^bIndicates that the patient currently has symptoms of depression or is receiving treatment for depression

^cDepression Anxiety and Stress Scale

^dPatients may have had more than one positive finding of multifidus dysfunction by MRI, PIT, MLT or aberrant movements

Patients had a long history of prior interventions including injections, ablations, physical therapy, and biologics reported as part of their ongoing care. Prior to enrollment, 58% of patients reported having had at least one epidural injection, 32% had at least one medial branch injection, and 21% had at least one facet joint injection. Additionally, 20% reported undergoing radio frequency ablation on at least one occasion. Table 5 shows access to these interventions during the randomization phase and indicates that fewer interventional approaches were required in the treatment arm relative to the control arm.

Table 5.

Number of patients receiving at least one treatment by category in the 12 months post-randomization

	Treatment (n = 94)	Control (n = 94)
Conservative treatments a	22.3% (21)	36.2% (34)
Ablation	0.0% (0)	9.6% (9)
Block/injection b	12.8% (12)	40.4% (38)
Psychological/behavioral c	5.3% (5)	10.6% (10)
Physiotherapy and exercise	31.9% (30)	47.9% (45)

^a Acupuncture, chiropractic, massage; ^b epidural, medial branch, and facet; ^c counseling, cognitive behavioral therapy

At baseline, both the treatment and control arms were using a similar profile of medication (Table 6), with 32% and 38% of patients in the treatment and control arms on opioid medication, respectively. As patients’ medication was optimized at baseline, sites were instructed to minimize changes to prescribed medications over the randomization period to avoid confounding the primary endpoint.

Table 6.

Baseline medication use

Medication	Treatment n = 99 Pt (% Pt)	Control n= 104 Pt (% Pt)	Total N = 203 Pt (% Pt)
At least one medication for LBP	91 (92%)	94 (90%)	185 (91%)
NSAID	58 (59%)	55 (53%)	113 (56%)
Opioid	32 (32%)	40 (38%)	72 (35%)
Simple analgesic	32 (32%)	31 (30%)	63 (31%)
Muscle relaxant	31 (31%)	44 (42%)	75 (37%)
Anticonvulsant	23 (23%)	16 (15%)	39 (19%)
Benzodiazepine	1 (1%)	5 (5%)	6 (3%)
Antidepressant	7 (7%)	7 (7%)	14 (7%)
COX-2 inhibitor	4 (4%)	2 (2%)	6 (3%)
Corticosteroid	0 (0%)	2 (2%)	2 (1%)
Local anesthetic	1 (1%)	2 (2%)	3 (1%)
Other	2 (2%)	4 (4%)	6 (3%)

COX-2 cyclooxygenase-2, LBP low back pain, NSAID nonsteroidal anti-inflammatory drugs

Primary Endpoint

The primary endpoint of the mean change in ODI between the treatment and control arms at the 1-year follow-up visit with MMRM for missing data (Table 7) was statistically significant, with a clinically relevant mean improvement in the treatment versus control arms: ODI (− 19.7 ± 1.4 vs. − 2.9 ± 1.4; between-group difference of 16.8 ± 1.9; 95% CI [− 20.6 to − 13.0 points]; p < 0.001).

Table 7.

One-year outcomes for primary, secondary using MMRM

	Baseline		Change from baseline		Difference in mean ± SE	p-value
	Treatment mean ± SE	Control mean ± SE	Treatment mean ± SE	Control mean ± SE
Primary endpoint cohort a	(n = 99)	(n = 104)
Primary outcome
ODI	44.1 ± 0.9	44.2 ± 0.8	− 19.7 ± 1.4	− 2.9 ± 1.4	− 16.8 ± 1.9	p < 0.001
Secondary endpoints
NRS	7.1 ± 0.1	7.1 ± 0.1	− 3.6 ± 0.2	− 0.6 ± 0.2	− 3.0 ± 0.3	p < 0.001
EQ-5D-5L	0.626 ± 0.012	0.612 ± 0.012	0.155 ± 0.012	0.008 ± 0.012	0.147 ± 0.018	p < 0.001

CI confidence interval, EQ-5D-5L EuroQol Five-Dimension Scale, MMRM mixed model for repeated measures, NRS numeric rating scale, ODI Oswestry Disability Index

^a Primary endpoint cohort with MMRM for missing data

Secondary Endpoints

The secondary endpoints for this trial are reported in Table 7. Using MMRM for missing data, all secondary endpoints were statistically significant and were greater than reported minimum clinically important differences for change in back pain (NRS) (−3.6 ± 0.2 vs. −0.6 ± 0.2; between-group difference of −3.0 ± 0.3; 95% CI [−3.6, −2.5]; p < 0.001) and for change in healthcare-related quality of life (EQ-5D-5L) (0.155 ± 0.012 vs. 0.008 ± 0.012; between-group difference of 0.147 ± 0.018; 95% CI [0.112, 0.1821]; p < 0.001) [37, 38].

A completer analysis (Table 8) of change from baseline in ODI (−20.0 ± 1.8 vs. −3.0 ± 1.3; between-group difference of −17.0; 95% CI [−21.5, −12.6]; p < 0.001), NRS (−3.5 ± 0.3 vs. −0.6 ± 0.2; between-group difference of −3.1; 95% CI [−3.7, −2.4]; P < 0.001), and EQ-5D-5L (0.157 ± 0.017 vs. 0.011 ± 0.015; between-group difference of 0.146; 95% CI [0.101, 0.191]; p < 0.001) (Fig. 2) showed a progressive improvement over time in patients treated with restorative neurostimulation in all three outcome measures, consistent with the restorative mechanism of action, while patients in the control arm did not derive durable or significant benefit over their baseline status.

Table 8.

Completer analysis of secondary endpoints

	Baseline		1 Year		Change From baseline		Difference in mean (95% CI)	p-value
	Treatment mean ± SE	Control mean ± SE	Treatment mean ± SE	Control mean ± SE	Treatment mean ± SE	Control mean ± SE
Completer analysis a	(n = 99)	(n = 104)	(n = 94)	(n = 94)
Secondary endpoints
ODI	44.1 ± 0.9	44.2 ± 0.8	24.1 ± 1.9	41.1 ± 1.4	−20.0 ± 1.8	−3.0 ± 1.3	−17.0 (−21.5, −12.6)	p < 0.001
NRS	7.1 ± 0.1	7.1 ± 0.1	3.5 ± 0.3	6.6 ± 0.2	−3.6 ± 0.3	−0.6 ± 0.2	−3.1 (−3.7, −2.4)	p < 0.001
EQ-5D-5L	0.626 ± 0.012	0.612 ± 0.012	0.783 ± 0.016	0.625 ± 0.015	0.157 ± 0.017	0.011 ± 0.015	0.146 (0.101, 0.191)	p < 0.001

CI confidence interval, EQ-5D-5L EuroQol Five-Dimension Scale, NRS numeric rating scale, ODI Oswestry Disability Index, SE standard error

^a Patients with baseline and 1-year outcomes

Fig. 2.

Mean changes over time for a mean Oswestry Disability Index for completers, b change in Oswestry disability index, c mean low back pain (NRS) for completers, d change in low back pain (NRS), e mean healthcare-related quality of life (EQ-5D-5L) for completers, and f change in healthcare-related quality of life (EQ-5D-5L). Differences between treatment and control were statistically significant for all follow-up time points (p < 0.001). EQ-5D-5L EuroQol Five-Dimension Scale, MMRM mixed model for repeated measures, NRS numeric rating scale, ODI Oswestry Disability Index

Tertiary Endpoints

The proportion of patients who reached the composite endpoint of ≥ 15-point ODI improvement and/or ≥ 50% NRS back and no worsening in either measure was 72% in the treatment arm and 11% in the control arm (p < 0.001).

The proportion of participants for whom global impression of change was “very much improved” or “much improved” was 89% in the treatment arm compared to 15% in the control arm (p < 0.001), and patients for whom treatment satisfaction was reported as “yes” rather than “no” or “maybe” was 76% versus 12% (p < 0.001) in favor of treatment (Table 9).

Table 9.

Completer analysis of tertiary endpoints

	1 Year		p-value
	Treatment mean ± SE (Min, Max) or Pt (% Pt)	Control mean ± SE (Min, Max) or Pt (% Pt)
Completer analysis a	(n = 94)	(n = 94)
Tertiary endpoints
NRS/ODI composite b	68 (72%)	11 (12%)	P < 0.001
PPR (%) c	62% ± 3.3	7% ± 1.8	P < 0.001
SGIC d	84 (89%)	14 (15%)	P < 0.001
Treatment satisfaction e	73 (78%)	3 (3%)	P < 0.001
Responder rate
≥ 50% NRS	50 (53%)	6 (6%)	P < 0.001
≥ 15-point ODI	66 (70%)	16 (17%)	P < 0.001

NRS numeric rating scale, ODI Oswestry Disability Index, PPR percent pain relief, SE standard error, SGIC subject global impression of change

^aPatients with baseline and 1-year outcomes, ^b≥ 15-point ODI improvement and/or ≥ 50% NRS back and no worsening in either measure, ^cpercent pain relief, ^dsubject global impression of change “very much improved” or “much improved,” ^e “yes” vs. “maybe/no”

The change in individual outcome measures (Fig. 3) shows that for treatment-arm patients, 70% achieved ≥ 30% and 53% achieved ≥ 50% improvement in NRS back, 73% achieved ≥ 30% and 52% achieved ≥ 50% benefit in ODI, and 48% achieved ≥ 30% improvement in EQ-5D-5L. In the control arm, 18% achieved ≥ 30% and 6% achieved ≥ 50% improvement in NRS back, 19% achieved ≥ 30% and 5% achieved ≥ 50% benefit in ODI, and 15% achieved ≥ 30% improvement in EQ-5D-5L. Pain remission, defined as a residual NRS of ≤ 3, was observed in 52% of treatment-arm patients and 6% of those in the control arm.

Fig. 3.

Individual patient percent change from baseline in a, b ODI, c, d back pain, and e, f EQ-5D-5L between treatment and control arms (red: patients with worse than baseline, white: patients with < 30% improvement, light green: patients with 30–50% improvement, and dark green: patients with ≥ 50% improvement), EQ-5D-5L EuroQol Five-Dimension Scale, NRS numeric rating scale, ODI Oswestry Disability Index

Safety Outcomes

The profile of related adverse events was similar to that in previously reported studies [22]. A total of 31 device-, procedure-, and/or therapy-related events occurred in 23 (23.2%) treatment-arm patients through 1-year follow-up. The most common events included implant site pocket pain occurring in eight (8.1%) patients, device overstimulation of tissue in five (5.1%) patients, lead fracture in three (3.0%) patients, and implant site seroma and implant site dermatitis in two patients (2%) for each. Additional events occurring once included anesthetic complications, device stimulation issues, delayed healing, implant site pocket infection, aggravated back pain, medical device site pain (not at pocket), radicular pain, shoulder pain, syncope, and wound infection. No lead migrations were observed. Eight system modifications were performed in seven patients (3 explants, 4 lead revisions, 1 implantable pulse generator [IPG] pocket revision; 1 patient had a lead revision followed by an explant). Of the 31 adverse events, 23 were resolved by 1 year, with pocket pain being the most common unresolved event in four patients.

Discussion

The objective of the RESTORE trial was to establish the effectiveness of restorative neurostimulation compared to OMM in the treatment of mechanical chronic low back pain due to multifidus dysfunction. The primary endpoint of 1-year change in ODI was statistically significant between the treatment arm and control arm, indicative of a clinically meaningful and durable effect. The prespecified secondary endpoints of reduction in pain and improvement in health-care-related quality of life were also statistically significant, demonstrating superiority and durability of restorative neurostimulation compared to OMM.

The RESTORE and the ReActiv8-B trials addressed similar cohorts. Patients enrolled in this trial presented with intractable chronic low back pain that had persisted for at least 6 months prior to enrollment (though on average greater than 10 years), resulting in pain more than 90% of days in the prior year, and had failed previous treatments, including pain medications and physical therapy. Additionally, the presence of multifidus dysfunction was confirmed by physical examination or radiological assessment. The results from this trial are consistent with the 1-year clinical outcomes reported in the ReActiv8-B trial (Table 10) [22]. The similarities in patient selection and 1-year clinical outcomes between the trials strongly suggest that the long-term durability seen in the ReActiv8-B trial [21] will also be replicated by patients in this trial. The safety profile was excellent compared to other neuromodulation procedures [39, 40] and consistent with previous studies of restorative neurostimulation [24, 41].

Table 10.

Comparative longitudinal outcomes with ReActiv8-B

	RESTORE (%, n/N)	1 Year ReActiv8-B [22]a (%, n/N, CI)
≥ 15-point ODI improvement	70 (66/94)	69 (121/176)
≥ 20-point ODI improvement	53 (50/94)	57 (101/176)
Remitter (resolution of back pain NRS ≤ 3 or VAS ≤ 2.5)b	52 (49/94)	52 (91/176)

CI confidence interval, NRS numeric rating scale, ODI Oswestry Disability Index, VAS visual analogue scale

^aCombination of control arm with only 8 months stimulation and treatment arm with 1 year stimulation,

^bRESTORE used the NRS pain scale and ReActiv8-B used VAS

The major differences between the ReActiv8-B and RESTORE trials were the control arms and the timing of the primary efficacy endpoint. In this trial, the control arm was ongoing OMM, whereas the ReActiv8-B trial employed sham stimulation as the control. The selection of OMM rather than sham as a control was a pragmatic decision. Following the Food and Drug Administration (FDA) approval and commercialization of the device, adequate blinding was deemed impossible, given the wide availability of patient-facing information reporting the sensations associated with strong muscle contractions that accompany restorative neurostimulation. Using a control arm of guideline-directed OMM adds important real-world context relating to the efficacy of currently available interventions and offers interesting insight into alternatives to current best practice for the management of CLBP.

These data on healthcare utilization (Table 5) also demonstrate a reduction in the application of interventional approaches in the treatment arm during the randomization phase compared to the control arm, and as the clinical benefit of this therapy takes time to accrue, we would expect to see further reductions over a longer time course. The comparison of annual utilization in the OMM group before and after crossover in subsequent publications will provide further detail.

Opioid use for managing mechanical CLBP is a complex and controversial topic, especially given the enormous financial and societal impact of the opioid epidemic in the United States. In this trial, 32% of patients in the treatment arm and 38% of patients in the control arm were on opioid medication at baseline. Patient medication was optimized at baseline, and sites were instructed to minimize dosing alterations as much as possible to avoid confounding the effect of other treatments at the 1-year primary endpoint. Longer-term follow-up and planned secondary analyses will further explore voluntary changes in medication associated with long-term symptom relief. We also predict meaningful reductions in medication use concordant with a reduction in symptom severity, similar to that observed in other studies [21].

Both the ReActiv8-B and RESTORE studies have demonstrated the safety, effectiveness, and long-term durability of restorative neurostimulation for a historically difficult-to-treat patient population. Given this comparison to OMM, it is appropriate to incorporate restorative neurostimulation in the CLBP care continuum. It is generally accepted that implanted devices should not be considered as a first-line treatment option, particularly for patients who have only recently developed back pain and may yet spontaneously improve. Noninvasive, conservative treatment should always be exhausted prior to considering something more invasive, and the advent of restorative neurostimulation does not change that. Nonetheless, the cost burden of repeated pain-mitigating spinal interventions for CLBP, the strictly palliative nature of those treatments, and the lack of viable, proven alternatives make a strong case for consideration of a definitive restorative intervention sooner rather than later. Patient identification is straightforward and highly selective of responders, as demonstrated by this trial and previous work. Potential candidates for restorative neurostimulation can be reliably selected on the basis of clinical history, imaging, and physical examination.

Finally, care should be taken in not misconstruing the implications of this trial. In no way should restorative neurostimulation be considered a panacea for all patients with chronic low back pain. On the contrary, the therapy is aimed at a very specific subset of CLBP patients with non-radicular, non-neuropathic pain, no indication for lumbar spine surgery and no previous lumbar spine surgery, a proven history of failure to respond to less invasive treatment, and perhaps most critically, clear clinical evidence for multifidus dysfunction.

Strengths and Weaknesses

This trial has some limitations. Participants in this trial were not blinded to their treatment, and as a result, those randomized to the control arm may have experienced a nocebo effect underestimating the clinical effect of OMM medical management. In addition, patients in the treatment arm may have experienced a placebo effect after being randomized to interventional treatment. Both of these effects were anticipated and contributed to the rationale for the timing of the 1-year primary endpoint. This timing allows for their impact to subside and for the full effects of OMM or restorative neurostimulation to accrue. The additional attention and monitoring afforded to patients in the treatment arm of this RCT were above standard management protocols for restorative neurostimulation. These additional clinical contact points may have resulted in consideration of additional interventions, artificially inflating healthcare utilization in the short term above what may typically be expected. The effect of longer follow-up on interventions will be reported in due course.

Conclusion

Therapies such as restorative neurostimulation that function via a novel mechanism of action require high-quality evidence to ensure that clinical adoption is both justified and aligned with best practices for patient safety and outcome. The RESTORE trial continues to demonstrate that this therapy is a safe, reversible, clinically effective, and highly durable option for patients with nonoperative CLBP associated with multifidus dysfunction. Based on the evidence generated from the RESTORE trial and from previous studies, restorative neurostimulation has more supportive high-quality data than any other neuromodulation approach for mechanical low back pain. This demonstration of treatment superiority over OMM through 1 year is a significant milestone in addressing a major health burden and unmet clinical need.

Author Contributions

Drs. Schwab, Gilligan, and Mekhail were involved in designing the trial; Dr. Klemme provided imaging analysis for patient inclusion; Drs. Langhorst, Heros, Gentile, Costandi, Moore, Gilmore, Manion, Chakravarthy, Meyer, Bundy, Tate, Sanders, Vaid, Szentirmai, Goree, V. Patel, Lehmen, Desai, Pope, Giuffrida, Hayek, Virk, Paicius, Mekhail, and Levy were involved in data acquisition; Drs. Schwab, Gilligan, Mekhail, K. Patel, Langhorst, Heros, Levy, and Chakravarthy were involved in analyzing and interpreting the data and preparing the first draft. All authors were involved in drafting/revising the paper and approving the final version of the manuscript, and they agree to be accountable for all aspects of the work. Drs. Schwab, K. Patel, Mehkail, and Gilligan were involved in trial oversight and patient selection.

Funding

Funding to support this trial was provided by Mainstay Medical, Inc. The Rapid Service and open access fees were paid for by Mainstay Medical, Inc.

Data Availability

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Conflict of Interest

Mainstay Medical (“Mainstay”) funded this trial and compensated all investigators and committee members either directly (consultant fees) or indirectly (payments to institution). Travel expenses related to investigator meetings and training were reimbursed only with prior authorization. Dr. Schwab was the principal investigator for the RESTORE Trial and reports consultant fees and royalties/licenses from Zimmer Biomet, Stryker, and Medtronic, consultant fees from Mainstay Medical and Medicrea, is an executive board member with International Spine Study Group, is a shareholder and has other interests with SeaSpine and VFT Solutions. Dr. K. Patel reports consultant fees from Abbott, Boston Scientific, Biotronik, Mainstay Medical, Vertos, Saluda, SPR Therapeutics, and fiduciary roles as Vice President for the North American Neuromodulation Society and Women Innovators in Pain Medicine, and Director-at-Large for the International Neuromodulation Society. Dr. Langhorst reports consultant fees, honoraria, and meeting support from Mainstay, consultant fees from Vivex, is an advisory/data safety monitoring board member for Vivex, and is a shareholder/holds stock/stock options in Boston Scientific and ATEC Spine. Dr. Heros reports grants, meeting support, and consultant fees from Saluda Medical, grants from Abbott and Ethos Laboratories, meeting support and consultant fees from Mainstay Medical, and consultant fees from Boston Scientific. Dr. Costandi reports grants from Vertos, Medtronic, Vivex, and Saluda. Dr. Moore reports honoraria from Mainstay Medical. Dr. Gilmore reports consultant fees from Mainstay, SPR Therapeutics, Nevro, Nalu, Biotronik, and Saluda, and other interests with SPR Therapeutics. Dr. Chakravarthy reports consultant fees and shareholder/stock options/stock from Mainstay. Dr. Tate reports consultant fees from Nevro, Saluda, Curonix, Vivex, Abbott, Medtronic, and Vertos, honoraria from the Virginia Pain Society and Georgia Society of Interventional Pain Practitioners, board memberships in Women Innovators in Pain Medicine and American Society for Pain and Neuroscience, and secretary of Georgia Society of Interventional Pain Practitioners. Dr. Sanders reports consultant fees and educational event support from Mainstay Medical and Vertos. Dr. Goree reports consultant fees from Saluda, Abbott, and Stratus Medical. Dr. V. Patel reports consultant fees from Mainstay, grants from Orthofix, Pfizer, Premia Spine, Medicrea, Globus, 3-Spine, and Spinal Kinetics, contracts and grants from Aesculap and Medtronic, contracts from Zimmer Biomet Spine, Inc., Johnson & Johnson Medical Device Business Services, NCS America, Simplify Medical, SI Bone, Orthobond Corp., Cerapedics, consultant fees from Spine Welding, SI Bone, expert testimony for Ogborn Mihm, LLP Expert Witness Deposition, and educational event support from Ecential Robotics and Johnson & Johnson Medical. Dr. Lehmen reports consultant fees, honoraria, and meeting support from Mainstay and Globus Medical-NuVasive. Dr. Desai reports royalties/licenses from Nevro, consultant fees from Medtronic, Nalu, and SPR Therapeutics, data safety monitoring/advisory board membership with FUSMobile, shareholder/stock/stock options in HypreVention, SPR Therapeutics, SynerFuse, and Virdio Health. Dr. Pope reports consulting fees, honoraria, research grants, educational support, and shareholder/stock options/stock from Abbott and Saluda, research grants, honoraria, royalties/license, consulting fees and shareholder/stock options/stock from Aurora, research grant from AIS, research grants, consulting fees, and honoraria from Boston Scientific, Flowonix, Ethos, Mainstay Medical, Muse, grants, consulting fees, honoraria, and shareholder/stock options/stock from Painteq, SPR Therapeutics, Theraquil, Vertos, consulting fees and honoraria from Medtronic, Tersera, Vertiflex, and WISE, consulting fees, honoraria, and shareholder/stock options/stock from Spark and SpineThera, shareholder/stock options/stock from Anesthetic Gas Reclamation, Axonics, Celeri Health, Pacific Research Institute, and Stimgenics, royalties/licenses and shareholder/stock options/stock from Neural Integrative Solutions, royalties/licenses from Elsevier and Springer, a patent for Neuronmonitoring, and fiduciary roles as Chairman for Pacific Spine and Pain Society and immediate past president of the American Society for Pain and Neuroscience. Dr. Giuffrida reports consultant fees, medical writing support and a medical advisory board membership from Mainstay, and executive board membership for the American Society of Pain and Neuroscience. Dr. Hayek reports payment for expert witness defense testimony, data safety/monitoring board membership on three non-industry funded studies, and was 2022–2023 past president of North American Neuromodulation Society. Dr. Virk reports research/educational event support from Mainstay, and consultant fees and research support from LifeSpine. Dr. Paicius reports consulting fees from Abbott and Biotronik; Dr. Mekhail functioned as independent medical monitor of the RESTORE trial has a consultancy agreement with Mainstay Medical. Dr. Levy reports unpaid consultancies with Abbott, Biotronik, Nalu, and Saluda, and stock options with Nalu and Saluda. Dr. Gilligan reports consultant fees from Mainstay Medical, Saluda, Persica, Iliad Lifesciences, and Biotronik, expert witness testimony payment, and fiduciary roles as Editor in Chief of Pain Practice, finance committee member for North American Neuromodulation Society, and on the board of directors for International Neuromodulation Society. Drs. Gentile, Bundy, Manion, and Klemme report consultant fees from Mainstay Medical. Drs. Meyer, Vaid, and Szentirmai have no other disclosures.

Ethical Approval

The RESTORE trial followed the principles of the Declaration of Helsinki of 1964 and its later amendments and Good Clinical Practice (GCP). The WCG IRB acted as the Central IRB (RN#20,211,219) for most sites while other sites had local IRB approval prior to enrollment of patients. All patients provided written informed consent to participate in the trial.