Study Design and Setting

This study was a prospective, unicentric, randomized, double-blinded, controlled trial conducted in a public hospital (The First Affiliated Hospital, Sun Yat-sen University) from May to December 2023.

Ethical Approval

The clinical research prospectively received approval from the ethics committee of the First Affiliated Hospital, Sun Yat-sen University ([2022]441), and was registered on ClinicalTrials.gov (ChiCTR2300070943). The research was conducted in accordance with the Declaration of Helsinki, and all patients provided informed consent.

Participants

A sample of 40 patients (27 females and 13 males) with CNLBP was recruited and randomly divided as 1:1 ratio into 2 groups: rPMS group (rPMS and CMT) or sham-rPMS group (sham-rPMS and CMT) (Fig. 1).

Fig. 1

Consolidated standards of reporting trials (CONSORT) flow diagram; rPMS repetitive peripheral magnetic stimulation, CMT core muscle training, VAS visual analog scale

The inclusion criteria were: (1) pain in the area below the T12 margin, above the transverse stria of the buttocks, and between the axillary midlines on both sides, with a duration of not less than 3 months, without numbness and radiative pain of the lower limbs, diagnosed with CNLBP based on International Classification of Diseases (ICD-10) code M54.5; (2) between 18 and 55 years of age; and (3) no pacemaker or implanted stent in the body. The exclusion criteria were: (1) specific low back pain or other chronic pathological pain caused by cancer/tumor, tuberculosis, fracture, infection, and inflammatory diseases; (2) associated with sciatica or root pain syndrome (e.g., positive sign in the Lasègue test, abnormal knee-jerk reflex, and Achilles tendon reflex, sensory disturbance, muscle strength decline); (3) definite orthopedic disease (e.g., spinal trauma, severe osteoporosis, related surgical history); (4) complicated with serious circulatory, respiratory, neurological or cardiac diseases; (5) previous history of epilepsy, craniocerebral injury, cerebrovascular disease, and craniocerebral surgery; (6) pregnant and lactating women; (7) obesity with body mass index (BMI) > 30; (8) under other analgesic medication or treatment, or training targeted the core muscle in the last 3 months; and (9) having undergone any medications that affect the postural control ability, muscle strength, and cortical excitability in the last 3 months.

Patients underwent generalized physical examinations (vital signs and preliminary examinations on respiratory, circulatory and neurology to avoid serious underlying diseases; specific examinations including palpation on tenderness point, percussion pain, Lasegue test, Bragard sign, Garslen test, manual muscle strength test, etc.) and brief history-taking before enrollment according to the assessment consensus of the Chinese Association of the Study of Pain [30], conducted by a certified professional physician with 5 years of clinical experience. If eligible, the participants signed the informed consent for study participation and entered the pre-intervention assessment stage after random allocation and received their first intervention the next day. To control the expectation bias, participants were told that two types of rPMS interventions adopted in this study had comparable treatment effects in CNLBP. Patients must complete the post-intervention assessment within 1–3 days after the last treatment to ensure accuracy and quality of the data.

InterventionsRepetitive Peripheral Magnetic Stimulation (rPMS)

A magnetic field stimulator with a figure 8 coil (CCY-IA; Wuhan Yiruide, Wuhan, Hubei, China) was used to provide stimulation on the lumbar (L4–L5) paravertebral region of the pain side (Supplementary Materials 1). The stimulation parameters [19, 20] consisted of a 5-s stimulation at a frequency of 20 Hz with an interval of 15 s. The stimulation intensity was 35–40% of the maximum stimulation output to provide a comfortable, painless, and palpable muscle contraction. The rPMS group received a total stimulus of 6000 pules during a 20-min session, once daily, 3 times per week for 4 consecutive weeks. The sham-rPMS was administrated by placing the coil vertically upside-down without altering the stimulation parameters, ensuring no palpable muscle contraction. All participants had no prior experience receiving peripheral magnetic stimulation (PMS)/transcranial magnetic stimulation (TMS).

Core Muscle Training (CMT)

The CMT therapy primarily focused on the isometric contraction of the bilateral deep lumbar multifidus (MF). An ultrasound imaging was performed on MF to provide real-time visual feedback, educating patients on the correct contraction technique before training. Patients were supervised one-on-one by a professional therapist who possess qualifications and have accumulated over 2 years of clinical experience, ensuring the provision of high-quality care. The treatment protocol comprised 12 sessions, each lasting approximately 30 min, which were further divided into two phases: isometric contraction of the transverse abdominis and lumbar MF (Supplementary Materials 1).

Transverse abdominis isometric contraction [10]: in a supine position with arms aside, patients were instructed to slowly contract their abdomen toward their spine. Next, extend their spine slowly by pressing against the cushion plate under their neck and sacrum. Each stretch lasted for 10 s and was repeated 10 times, with a 30-s rest in between.

Bird-dog exercise [10]: in a quadruped position with straightened forearms, patients were instructed to maintain the stability of their head and trunk while bending one shoulder forward to 90°, and extending the opposite hip joint backward to 60°–90°. Each lasted for 10 s and was repeated 10 times on each side, with 3 sets and a 30-s rest in between. Incorrect posture, such as a concave waist or a convex prothorax, should be avoided to prevent spinal muscle injury. This exercise mainly involves lumbar MF isometric contraction.

OutcomesPrimary Outcome MeasuresPain Intensity and Disability Severity

Research has shown that the visual analog scale (VAS) demonstrates good reliability and validity in assessing the level of pain [31]. Mark 0 (painless) and 10 (most painful) on both ends of a 10-cm straight line, ask the patient to mark their current pain level. The Oswestry Dysfunction Index (ODI) is a reliable and valid measure to self-quantify the dysfunction [32]. There are 10 daily function-related questions. Converted the total score to percentages, 0% represents normal, and, the closer to 100%, the more serious the dysfunction. There is one sexual life-related question, which can be left blank for those without sexual life experience and will not be included in the calculation. The minimal clinically important difference (MCID) for VAS and ODI was set as a 30% improvement relative to the baseline [33].

Postural Control Performances

Patients were asked to perform four balancing tasks (eyes-opened/closed standing on a stable/unstable plane) while wearing a portable functional near-infrared spectroscopy (fNIRS) device on a force balance instrument (PK254P; TecnoBody, Bergamo, Italy) (Fig. 2a). Before the commencement of the test, the patients were thoroughly informed of the posture requirements for each task and practiced 1–2 times to ensure a complete understanding of the instructions. The balancing platform can achieve a stable (static balancing task) or unstable (dynamic balancing task) plane by adjusting its softness or hardness. The patients were asked to stand on both feet at the center of the force plate (feet position standardized by a V-shape frame) with arms aside. A 10-s eyes-opened standing on the stable plane was set as the baseline. A 30-s task and a 30-s rest (baseline posture) were performed for three trials (Fig. 2c). Participants were instructed to concentrate on their balance control during the tasks while mentally relaxing during the rest and baseline intervals. The sway area (mm2) of the COP and the anterior–posterior (AP)/medial–lateral (ML) velocity (mm/s) were collected during the static and dynamic balancing tasks. Meta-analysis showed that the sway area showed good reliability and validity [34].

Fig. 2

Functional near-infrared spectroscopy (fNIRS) data were acquired during the balance test. a The patient performed various dynamic and static postural tasks on a pressure balance instrument, while fNIRS data was being collected simultaneously. b The distribution of the 6 regions of interest (ROIs). c The block design of the balance tasks. This involved a 10-s baseline (standing on a stable plane with eyes-opened), followed by 3 trials. Each block comprised a 30-s task and a 30-s rest period. The tasks included standing with eyes-opened/closed on a stable/unstable plane. The rest period involved maintaining the same posture as the baseline. PFC prefrontal cortex, SMA supplementary motor area, M1 primary motor cortex

Secondary Outcome MeasuresfNIRS Acquired During the Balancing Tasks

The fNIRS equipment (NirSmart-6000A; Danyang Huichuang Medical Equipment, Jiangsu, China) consists of 14 light source sensors and 14 detection sensors, forming 35 effective channels with a 30-mm distance in between, covering the bilateral side of the PFC and the motor region (Fig. 2b). The sensors are fixed by a soft head cap, providing a better adherent to the scalp to improve light coupling efficiency and signal strength [35]. The front edge of the cap is horizontal to and above the eyebrow, while the vertical axis lines on the midline of the head. The placement of sensors was according to the configuration of the standard international 10/20 system and performed on a standard adult Brodmann Talairach brain model. The Montreal Neurological Institute coordinates and the percentage of overlapped brain regions of each channel were then calculated [36]. D11 is located at the C3 position, S7 is located at the C4 position, and D3 is located at the Fpz position [8]. The concentration change of oxygenated hemoglobin (ΔHbO) and deoxyhemoglobin (ΔHbR) in various cortical regions were collected at a sampling rate of 11 Hz, with a wavelength of 730 nm and 850 nm, respectively. The consensus among experts indicates that fNIRS demonstrates robust reliability and validity when utilized for the assessment of gait or postural control [37].

Data Processing

Use the built-in analyzing software (NirSpark; Danyang Huichuang Medical Equipment) for data preprocessing. Firstly, the original optical density data were corrected for motion artifacts and then filtered using 0.01–0.2 Hz band-pass filtering to eliminate baseline drift or high-frequency cardiac or respiratory activities. Secondly, based on the modified Beers–Lambert law, the filtered optical density data was converted into ΔHbO. Thirdly, the block average of HbO of the three trials was calculated. The above pretreatment procedures were performed according to the previous studies [8, 38, 39]. We focused on the ΔHbO for the subsequent statistical analysis as it is more sensitive to hemodynamics [40].

The general linear model (GLM) was used to analyze the functional timeline of ΔHbO data for block-related designs. The data were regressed to a typical hemodynamic response function value (HRF) that mimics the actual hemodynamic response [41]. By fitting the experimental HRF with the block-designed typical HRF, an activation correlation coefficient, known as the β value, was calculated. This represents the activation intensity of the cerebral cortex during a specific task [42]. The higher the value, the greater the activation level of the brain region. We combined 6 postural control-related ROIs based on the anatomical location and corresponding channel information of the brain region: left and right PFC (BA9, 10, 11, 45, 46, 47), left and right M1 (BA 4), and left and right SMA (BA6), each of which contained 1–7 valid channels. Group analysis was performed according to the mean β value of each ROI.

Sample Size Calculation

Based on the analysis of our previous pilot study (n = 3 per group), G*power software (Franz Faul, Kiel University, Kiel, Germany) was used to estimate the sample size in terms of pain intensity (VAS). With an assumed effect size of 0.370, a correlation coefficient of 0.234, an α level of 0.05, and a power of 0.80, we determined that a minimum of 38 patients was required. To accommodate a possible dropout rate of 5%, we enrolled a total of 40 participants for this study.

Randomization and Blinding

The random grouping numbers generated in the SPSS 22.0 statistical software were coded as "A" and "B" by a third party who did not participate in the study. The opaque envelope with the grouping scheme was sealed and coded in sequence. According to the enrollment sequence of patients, the therapist unfolded the envelope to provide corresponding intervention. Both the patients and the researchers conducting data collection and statistical analysis were unaware of the grouping. The grouping codes were disclosed after completing the post-intervention assessment for preliminary statistical analysis. The corresponding treatment scheme was revealed after the data statistics as the final unblinding.

Statistical Analysis

The statistical analysis in this study was conducted by SPSS 22.0 (IBM, Somers, USA). The homogeneity of variances and normality of the distribution were determined using Levene's and Shapiro–Wilk tests, respectively. Continuous variables were analyzed using the t test or Wilcoxon rank-sum test, while categorical data was analyzed using the chi-square test or Fisher's exact test. To assess significant differences between treatments and within each group pre- and post-intervention, a two-way repeated measure ANOVA (time × group) was applied to determine the presence of interaction effects. In the fNIRS data analysis, a three-way repeated measure ANOVA (task × time × group) was performed to determine the significant interaction between four tasks (stable eyes-opened, stable eyes-closed, unstable eyes-opened, unstable eyes-closed), two times (pre- and post-intervention), and two groups (rPMS and sham-rPMS). If a significant interaction appeared, a post hoc analysis with a Bonferroni adjustment was carried out for multiple comparisons. Spearman's correlation was used to determine the association between variables. A statistically significant difference was represented by p < 0.050.