Lumbar stabilization exercise with and without real-time ultrasound imaging biofeedback in chronic low back pain patients: a randomized controlled trial

Abstract

Reduced lumbar multifidus (LM) cross-sectional area (CSA) has been linked with non-specific chronic low back pain (NCLBP); hence, lumbar stabilization exercise (LSE) is typically used as treatment. Real-time ultrasound imaging (RUSI) biofeedback can enhance performance and retention in LM activation. This study aimed to compare the effects of LSE with or without RUSI-biofeedback in patients with NCLBP. A single-blind, three-arm, parallel randomized controlled trial involving 90 participants, randomized into LSE_{RUSI-biofeedback,} LSE_{no-biofeedback}, or minimal intervention (n = 30 per group) was conducted. All interventions were administered twice weekly for 8 weeks, and outcomes (LM-CSA, pain, disability, and quality of life) were evaluated before and at 8-week and 20-week follow-ups. Compared to LSE_{no-biofeedback} and minimal intervention, LSE_{RUSI-biofeedback} demonstrated a greater increase (p < 0.05) in LM-CSA at 8 weeks (mean [95%CI] difference: 0.81 [0–1.62] and 1.79 [0.91–2.65], respectively) and at 20 weeks (1.61 [0.79–2.42] and 2.66 [1.79–2.53], respectively). LSE_{RUSI-biofeedback} showed a greater improvement (p < 0.05) in mental health scores at 8 weeks (3.63 [0.32–6.94]) and 20 weeks (7.34 [4.03–10.60]) compared to LSE_{no-biofeedback}. However, LSE_{no-biofeedback} showed a greater physical health scores at 20 weeks (− 4.89 [− 7.92 to − 1.86]) compared to LSE_{RUSI-biofeedback}. No significant differences (p > 0.05) were found between LSE_{RUSI-biofeedback} and LSE_{no-biofeedback} for pain or disability. The increase in LM-CSA significantly correlated with changes in pain (r = − 0.40 to − 0.50, p < 0.05) and disability (r = − 0.38 to − 0.42, p < 0.05). In conclusion, LSE with or without RUSI-biofeedback improved LM-CSA, pain, disability, and quality of life. However, LSE_{RUSI-biofeedback} led to a greater increase in LM-CSA. Such an increase appears to be related to changes in pain and disability. Trial registration: The study was registered in the Pan African Clinical Trials Registry (16/01/2018; PACTR201801002980602).

Introduction

Low back pain (LBP) is a highly prevalent musculoskeletal disorder, and a major global public health concern¹. Over the past three decades, it has been the leading cause of disability globally², leading to reduced quality of life (QOL) and imposing an enormous clinical and economic burden^3,4. Approximately 619 million people globally have LBP in 2020, and 843 million will be affected by 2050². Given the momentous burden of this disorder in addition to its high recurrence rate, evidence-based, optimized treatments are highly desirable³.

Although 90–95% cases of LBP are non-specific⁵, several factors contributing to pain have been proposed⁶, with muscular dysfunction, particularly in the deep stabilizing muscles of the spine, playing a critical role⁷. Among these muscles, the lumbar multifidus (LM), which is essential for maintaining spinal stability, is commonly implicated⁸. Specifically, alterations in the morphology of this muscle, including atrophy and reduced cross-sectional area (CSA)^9–11 have been documented in patients with non-specific chronic LBP (NCLBP) and suggested as a possible reason for its persistence⁸.

Traditional rehabilitation approaches for NCLBP often include therapeutic exercises targeting the trunk musculature. However, general exercise may not effectively address LM dysfunction¹². Thus, lumbar stabilization exercise (LSE), designed to train LM, has gained increasing attention as a targeted intervention¹³. This exercise focuses on improving neuromuscular control and stability, which are vital for ameliorating back pain symptoms^14,15.

The integration of real-time ultrasound imaging (RUSI) as biofeedback during re-training of muscle contraction has emerged as a novel approach¹⁶. RUSI provides visual feedback, allowing individuals to observe the contraction and activation of their deep trunk muscles in real-time^17,18. Given that LSE is based on motor control principles designed to enhance sensorimotor integration (i.e., use sensory input to generate appropriate motor output), RUSI can be used to augment the sensory input to generate more accurate motor output¹⁹. Importantly, experimental studies on asymptomatic individuals have shown the utility of the RUSI in augmenting activation performance and retention in the ability to activate the deep trunk muscles, including LM^19,20. Thus, enhancing precise trunk muscle activation and control could lead to better pain reduction and improved function²¹.

Despite the potential benefits of integrating RUSI biofeedback with LSE, limited research substantiates its specific effects, particularly on LM CSA in patients with NCLBP. A previous randomized controlled trial (RCT) showed that LSE is as effective as general exercise for improving LM thickness and reducing disability²². Another RCT comparing LSE against two other general back extensor exercise regimens showed positive results for LM CSA increase in favor of one of the general exercise approaches²³. While these trials contradict the notion that general exercise would be inadequate to restore LM dimensions, none of them integrates RUSI biofeedback in the LSE program. In an earlier trial²⁴ exploring the effects of LSE with RUSI biofeedback compared to standardized medical care (analgesics and advice) for acute LBP patients with between-sides asymmetry in LM CSA of > 11%, long-term follow-up at 1 and 3 years showed a link between improvement in LM CSA and reduced pain recurrence in favor of LSE. Similarly, a study²⁵ with a pretest–posttest design showed improved LM CSA and pain scores among cricketers with current or previous LBP who received LSE training with RUSI, compared with cricketers without LBP who did not receive LSE. Also, our pilot study²⁶ showed significant improvements in LM CSA, pain, and disability scores in patients with NCLBP receiving LSE with RUSI biofeedback, paving the way for a large-scale powered RCT to elucidate these findings.

To the best of our knowledge, no RCT has tested the assertion that LSE training with RUSI biofeedback has added benefits in a sample of patients with NCLBP using LM CSA, pain, disability, and QOL as outcomes. Moreover, there is limited evidence to suggest that post-LSE changes in LM size may be related to changes in pain and disability²¹. Therefore, the present study was conducted to evaluate the effects of LSE training with RUSI biofeedback on these key outcomes, providing valuable insights into the role of targeted rehabilitation for NCLBP. Our experimental hypothesis was that LSE training with RUSI biofeedback would be more effective at improving LM CSA, pain intensity, functional disability, and QOL compared to LSE training without RUSI biofeedback or minimal intervention. We also hypothesized that post-intervention change in LM CSA would be associated with post-intervention changes in pain intensity and functional disability.

Materials and methods

Study design

This was a prospective single-blind, three-arm RCT. This trial followed the Consolidated Standards of Reporting Trials Statement (CONSORT)²⁷, and a detailed protocol has been published previously²⁸, with no significant changes made in the present study that are different from the protocol.

Study settings

This study was conducted at the physiotherapy department, National Orthopaedic Hospital, Dala, Kano State, Nigeria, between March 2018 and September 2021.

Ethical consideration

The Research and Ethical Committee of National Orthopedic Hospital, Dala, Kano, approved this study (Ref: NOHD/RET/ETHIC/60). Written informed consent was obtained from each participant before inclusion in the study. All procedures conformed to the Helsinki Declaration. The study was registered in the Pan African Clinical Trials Registry (16/01/2018; PACTR201801002980602).

Participants’ selection

The participants were recruited at the physiotherapy outpatient clinic, National Orthopaedic Hospital Dala, through referral from a general practitioner or orthopaedist. The inclusion criteria were: (1) male or female aged 18 to 60 years, (2) LBP with or without leg pain of ≥ 12 weeks, and (3) ability to read and understand Hausa, English, or both. The exclusion criteria were: (1) history of spine surgery, (2) obvious deformities affecting the trunk or extremities, (3) serious spinal pathology (e.g., infection, fracture, metastases, cauda equina syndrome), (4) history of medical or surgical conditions that could hinder exercise performance, and (5) pregnancy or lactation.

Sample size calculation

The sample size was calculated to detect a minimum difference of 2.0 points in LM CSA between the LSE_{RUSI-biofeedback} and minimal intervention group and 1.0 point between LSE_{RUSI-biofeedback} and LSE_{no-biofeedback}, assuming a common standard deviation of 2.5 points based on our pilot study²⁶. Using an effect size of 0.32, an alpha of 0.05, a statistical power of 0.80, an attrition rate of 35%, and a correlation of 0.5 among repeated measures, for repeated measures, between factors ANOVA, a total of 90 participants (30 per group) were required.

Randomization and blinding

After obtaining signed informed consent and assessing baseline demographic and clinical characteristics as well as self-report questionnaires, a record officer, who was not involved in other study aspects, prepared consecutively numbered, sealed opaque envelopes using a computer-generated randomization sequence. This blinded data assessors to ensure unbiased outcome assessment. Permuted blocks of varying sizes (3 and 6) were used to maintain an equal number of participants in each group. The participants were randomly allocated into one of three groups (i.e., LSE_{RUSI-biofeedback}, LSE_{no-biofeedback}, minimal intervention [control]), of equal size (n = 30). Due to the nature of the intervention, blinding of the physiotherapist responsible for providing treatment was not possible.

Outcome assessment

Self-report questionnaires for pain intensity, functional disability, and QOL were assessed by a physiotherapist with over 3 years of clinical experience, while LM CSA was assessed by a consultant radiologist with over 10 years of diagnostic imaging experience. Both were blinded to the study conditions and patient group allocation. All outcomes were assessed before (baseline) and at 8 weeks (immediately) and 20 weeks after intervention.

Primary outcome

The primary outcome was the LM CSA measured using a RUSI machine (Edan D3 version 1.6, China) with a 5-MHz curvilinear transducer. Participants lay in a prone position on a plinth with arms by their sides and head turned to their preferred side. A pillow was placed under the hips to maintain a neutral lumbar spine (lumbosacral angle of < 15°). The L5 spinous process was palpated and marked for reference, and an acoustic coupling gel was applied. The transducer was placed longitudinally to confirm the L5 level, then transversely over the L5 spinous process to capture images of the spinous process, laminae, and LM on both sides^26,29. Transverse images of the LM CSA were obtained bilaterally, with the smaller side considered the LM CSA. Measurements were taken during maximal voluntary isometric contraction, and average values from three images were used for accuracy. Higher values indicate better CSA. Our pilot study showed a high intraclass correlation coefficient of 0.96 for this measurement.

Secondary outcomes

The secondary outcomes were pain intensity, functional disability, and QOL assessed using self-report questionnaires as described in detail in the protocol paper²⁸. Pain intensity was assessed using the Numerical Pain Rating Scale (NPRS, 0–10 cm). Both the English³⁰ and Hausa³¹ versions were used and have been shown to be reliable, valid, and responsive^30–32. Functional disability was assessed using the Roland-Morris Disability Questionnaire (RMDQ, 0–24), where 0 indicates no disability and 24 indicates maximum disability³³. Both the English³³ and Hausa³⁴versions were used and have been shown to be reliable and valid^33,34. For QOL, the 12-item Short-Form Health Survey (SF-12)³⁵ was used to assess physical and mental health, using the physical component scores (PCS-12, 0–100%) and mental component scores (MCS-12, 0–100%), respectively, with higher scores indicating better QOL. Both the English³⁵ and Hausa³⁶ versions were used, and both are reliable and valid^35,36.

Interventions

The study interventions began immediately following randomization. All interventions were provided twice a week for 8 weeks. The primary investigator, with over 15 years of clinical experience, applied LSE in the two experimental groups, while another therapist with over 6 years of clinical experience applied the minimal intervention. Participants in the LSE_{RUSI-biofeedback} and LSE_{no-biofeedback} groups also received the intervention provided to the minimal intervention group. All participants were advised against engaging in any exercise-related interventions during the study. They were encouraged to perform group-specific exercises at least twice daily at home and to continue these exercises after the 8-week treatment. Details of the interventions can be found in the published protocol paper²⁸.

LSE_{no-biofeedback} group

Initially, the participants were educated using videos and images on the lumbar spine anatomy and key stabilizing muscles, focusing on LM. They were then taught active isometric contraction of the LM, in synergy with other key deep stability muscles such as transverse abdominis (TrA), while lying prone with a pillow under their hips for neutral spine positioning. Contractions were held for 10 s, repeated 10 times, with 30- to 60-s rest intervals. Exercise progression was achieved by increasing the duration of contractions from 10 to 15 s by the 4^th week.

LSE_{RUSI-biofeedback} group

The LSE program was similar to that described for LSE_{no-biofeedback} group, with the addition of RUSI biofeedback to enhance LM contraction precision. While maintaining a neutral spine in a prone lying, the ultrasound transducer was placed transversally over the L2–L5 levels. Participants were instructed to perform isometric contractions of the LM in synergy with other deep stabilizing muscles, focusing on the monitor to track changes in LM thickness and maximize effort with each contraction (Fig. 1A–D). Contractions were held for 10 s, repeated 10 times, with 30- to 60-s rest intervals. Exercise progression was achieved by increasing the duration of contractions from 10 to 15 s by the 4^th week.

Fig. 1.

(A) LSE training with RUSI biofeedback (B) LM CSA imaging before treatment in the LSE_{RUSI-biofeedback group} (C) LM CSA imaging at 8-week follow-up in the LSE_{RUSI-biofeedback group} (D) LM CSA imaging at 20-week follow-up in the LSE_{RUSI-biofeedback group}. CSA, cross-sectional area; LM, lumbar multifidus; LSE, lumbar stabilization exercise; RUSI, real-time ultrasound imaging.

Minimal intervention (control) group

Three exercises commonly prescribed for chronic LBP disorders were performed. They include knee to chest, lumbar rotation, and bridging. Each exercise was held for 10 s, repeated 10 times, with 5- to 10-s rest intervals between repetitions. Exercise progression was achieved by increasing the duration of contractions from 10 to 15 s by the 4^th week.

Statistical analysis

Intention-to-treat (ITT) analysis was performed, including all randomized participants with available outcome data, regardless of follow-up. Normality of the dependent variables was tested using the Kolmogorov–Smirnov test. Descriptive statistics were used to summarize the data. One-way ANOVA was used to compare baseline continuous variables, while the Chi-square or Fisher’s exact test was used to compare baseline categorical variables across the three groups. To analyze the effects of treatment, a linear mixed-effects model was used by including time (baseline, 8 weeks, 20 weeks), treatment (LSE_{RUSI-biofeedback}, LSE_{no-biofeedback}, minimal intervention), and a time × treatment interaction as fixed effects, and a random intercept to account for within-subject correlations. Treatment effect was defined as the adjusted between-group difference and the associated 95% confidence intervals (CI). Post-hoc analysis was performed with Bonferroni correction for the primary outcome and Fisher’s least significant difference (LSD) for the secondary outcomes. Effect sizes were computed and interpreted using Cohen’s d criteria as trivial (< 0.2), small (≥ 0.2 and < 0.5), moderate (≥ 0.5 and < 0.8), and large (≥ 0.8)³⁷. The Pearson correlation coefficient was used to examine the relationship between post-treatment mean change scores in LM CSA and post-treatment mean change scores in pain intensity and functional disability in the experimental groups. Pearson correlation coefficients were interpreted as very weak < 0.20), weak (0.20–0.39), moderate (0.40–0.59), strong (0.60–0.79), and very strong (> 0.80)^38,39. All analyses were conducted using IBM SPSS Statistics version 23.0 (IBM Co., Armonk, NY, USA), at an alpha level of 0.05.

Results

Participation flow and sample characteristics

The flow of the participants through the study is depicted in Fig. 2. Of the 117 individuals with chronic LBP screened for eligibility, 90 (49 males [54.4%] and 41 females [45.6%]) with a mean (SD) age of 46.3 (11.2) years fulfilled the study criteria and were recruited. Most participants had a secondary school education (36%) and worked as civil servants (29%). During the 8-week intervention period, 5 participants withdrew from the study (LSE_{RUSI-biofeedback} group = 3, 10.0%; LSE_{no-biofeedback} group = 1, 3.3%; minimal intervention group = 1, 3.3%). An additional 16 participants withdrew during the 20-week follow-up (LSE_{RUSI-biofeedback} group = 3, 11.1%; LSE_{no-biofeedback} group = 4, 13.7%; minimal intervention group = 9, 31.0%) (Fig. 2). Dropouts were primarily due to loss of contact, illness, or personal issues. As shown in Table 1, there were no significant differences between the three groups in most of the baseline participant characteristics and outcome measures (p > 0.05). The groups differed significantly in education and occupational status, duration of LBP, and mental health (p < 0.05). No serious adverse effects were reported in relation to the interventions. However, 5 participants in the LSE_{no-biofeedback} group and 2 in the minimal intervention group reported temporary exacerbation of pain. Additionally, increased pain due to pre-existing musculoskeletal conditions, such as knee osteoarthritis, was reported by 4 participants in the LSE_{RUSI-biofeedback} group and 1 in the LSE_{no-biofeedback} group, and 3 in the minimal intervention group.

Fig. 2.

Flow of participants through the study.

Table 1.

Baseline demographic and clinical characteristics of the study participants.

Variables	LSERUSI-biofeedback (n = 30)	LSEno-biofeedback (n = 30)	Minimal intervention (n = 30)	p-value	Total (n = 90)
Age (years)	44.5 (10.9)	43.1 (10.0)	47.9 (8.53)	0.161a	45.2 (9.99)
BMI (kg/m2)	29.0 (5.64)	26.3 (4.47)	26.2 (4.74)	0.059a	27.2 (5.08)
Gender, n (%)
Male	20 (66.7)	13 (43.3)	16 (53.3)	0.191b	49 (54.4)
Female	10 (33.3)	17 (56.7)	14 (46.7)		41 (45.6)
Educational status, n (%)
Non-formal education	2 (6.7)	0 (0.0)	0 (0.0)	0.002**c	2 (2.2)
Primary education	7 (23.3)	3 (10.0)	15 (50.0)		25 (27.8)
Secondary education	12 (40.0)	12 (40.0)	12 (40.0)		36 (40.0)
Tertiary education	9 (30.0)	15 (50.0)	3 (10.0)		27 (30.0)
Occupational status, n (%)
Self-employed	13 (43.3)	9 (30.0)	11 (5.0)	0.041*c	33 (36.7)
Civil servant	7 (23.3)	14 (46.7)	8 (26.7)		29 (32.2)
Student	0 (0.0)	3 (10.0)	0 (0.0)		3 (3.3)
Housewife	10 (33.3)	4 (13.3)	11 (36.7)		25 (27.8)
Low back pain duration (years)	3.28 (0.89)	3.33 (1.84)	2.53 (0.62)	0.002a	3.05 (0.99)
LM CSA	5.63 (1.43)	5.57 (1.16)	5.94 (1.28)	0.514a	–
NPRS (0–10)	6.76 (0.72)	6.66 (0.60)	6.76 (0.72)	0.811a	–
RMDQ (0–24)	13.7 (1.88)	12.9 (2.68)	12.5 (2.27)	0.116a	–
PCS-12 (0–100)	30.1 (5.12)	30.1 (4.96)	30.3 (4.71)	0.964a	–
MCS-12 (0–100)	39.7 (3.83)	41.1 (5.67)	40.7 (7.12)	0.013a	–

LSE, lumbar stabilization exercise; BMI, body mass index; CSA, cross-sectional area; LM, lumbar multifidus; NPRS, Numerical Pain Rating Scale; RMDQ, Rolland-Morris Disability Questionnaire; PCS, physical component summary; MCS, mental component summary.

^aAnalyzed by one-way ANOVA.

^bAnalyzed by Chi-square test.

^cAnalyzed by Fisher’s exact test.

*p < 0.05.

**p < 0.01.

Results on LM CSA, pain intensity, functional disability, and QOL

For LM CSA, there was significant group x time interaction effect (F[4, 174] = 37.87, p < 0.001) and main effect of group (F[2, 87] = 9.742, p < 0.001) (Table 2). Post-hoc analysis with Bonferroni correction showed a significantly greater improvement in the LSE_{RUSI-biofeedback} group compared with the LSE_{no-biofeedback} and minimal intervention groups by an additional 0.81 points (95% CI, 0.00 to 1.62; p = 0.048; effect size = 0.58) and 1.79 points (95% CI, 0.91 to 2.65; p < 0.001; effect size = 1.10), respectively at 8-week follow-up. This superior improvement was sustained at 20-week follow-up by an additional 1.61 points (95% CI, 0.79 to 2.42; p < 0.001; effect size = 1.14) and 2.66 points (95% CI, 1.79 to 2.53; p < 0.001; effect size = 1.98), respectively. The LSE_{no-biofeedback} group was superior to the minimal intervention at all time points (p < 0.001). The main effect of time was also significant (F[2, 174] = 15.50, p < 0.001), with both the LSE_{RUSI-biofeedback} and LSE_{no-biofeedback} groups showing an increase in LM CSA over time (p < 0.001), except for the LSE_{no-biofeedback} group at 20-week follow-up (p = 0.721). In contrast, the minimal intervention group showed a decrease in the mean LM CSA over time (p < 0.001), indicating deterioration (Fig. 3).

Table 2.

Comparisons of LM CSA, pain intensity, functional disability, and QOL among the study groups based on ITT.

Outcome	Treatment group Mean (SD)			Adjusted between-group differences Mean (95% CI)
	LSERUSI-biofeedback	LSEno-biofeedback	Min intervention	LSERUSI-biofeedback vs LSEno-biofeedback	Cohen’s d	LSERUSI-biofeedback vs Min intervention	Cohen’s d	LSEno-biofeedback vs Min intervention	Cohen’s d	p-value†
LM CSA										< 0.001
Baseline	5.63 (1.43)	5.57 (1.16)	5.94 (1.29)	–		–		–
8 weeks	6.95 (1.74)	6.14 (0.93)	5.25 (1.33)	0.81 (0.00 to 1.62)*	0.58	1.79 (0.91 to 2.65)***	1.10	0.98 (0.13 to 1.82)*	0.78
20 weeks	6.98 (1.77)	5.37 (0.94)	4.41 (0.49)	1.61 (0.79 to 2.42)***	1.14	2.66 (1.79 to 2.53)***	1.98	1.06 (0.21 to 1.95)**	1.28
NPRS (0–10)										< 0.001
Baseline	6.76 (0.72)	6.66 (0.60)	6.76 (0.72)	–		–		–
8 weeks	2.73 (1.63)	2.23 (0.77)	6.23 (0.67)	0.50 (− 0.01 to 1.10)	0.39	− 3.56 (− 4.01 to − 2.98)***	2.81	− 4.00 (− 4.51 to − 3.45)***	5.54
20 weeks	2.86 (1.61)	3.16 (0.87)	6.16 (0.79)	− 0.30 (− 0.81 to 0.21)	0.23	− 3.30 (− 3.81 to − 2.78)***	2.65	− 3.00 (− 3.51 to − 2.48)***	3.61
RMDQ (0–24)										< 0.001
Baseline	13.7 (1.88)	12.9 (2.68)	12.5 (2.27)	–		–		–
8 weeks	5.33 (2.53)	5.00 (1.25)	12.4 (1.35)	0.33 (− 0.71 to 1.37)	0.17	− 7.10 (− 8.14 to − 6.05)***	3.49	− 7.43 (− 8.47 to − 6.38)***	5.69
20 weeks	4.40 (2.90)	4.93 (1.14)	12.1 (1.55)	− 0.53 (− 1.57 to 0.51)	0.24	− 7.77 (− 8.81 to − 6.72)***	3.31	− 7.23 (− 8.27 to − 6.18)***	5.27
PCS-12 (0–100)										< 0.001
Baseline	30.1 (5.12)	30.3 (4.71)	29.9 (5.20)	–		–		–
8 weeks	46.7 (7.66)	47.4 (4.50)	27.3 (4.34)	− 0.68 (− 3.72 to 2.34)	0.11	19.3 (16.2 to 22.3)***	3.11	20.0 (16.9 to 23.0)***	4.55
20 weeks	44.3 (6.69)	49.2 (5.04)	34.1 (8.73)	− 4.89 (− 7.92 to − 1.86)*	0.83	10.2 (7.20 to 13.2)***	1.31	15.1 (12.1 to 18.1)***	2.12
MCS-12 (0–100)										< 0.001
Baseline	39.7 (3.83)	40.7 (7.12)	43.8 (4.90)	–		–		–
8 weeks	56.3 (7.56)	52.7 (3.59)	33.1 (6.78)	3.63 (0.32 to 6.94)*	0.61	23.2 (19.9 to 26.5)***	3.23	19.6 (16.2 to 22.9)***	3.62
20 weeks	56.4 (7.56)	49.0 (4.63)	33.2 (9.78)	7.34 (4.03 to 10.6)***	1.18	23.1 (19.8 to 26.4)***	2.65	15.8 (12.5 to 19.1)***	2.07

CI, confidence interval; LSE, lumbar stabilization exercise; Min, minimal; BMI, body mass index; CSA, cross-sectional area; LM, lumbar multifidus; NPRS, Numerical Pain Rating Scale; RMDQ, Rolland-Morris disability questionnaire; PCS, physical component summary; MCS, mental component summary.

Higher scores in LMM CSA, PCS-12 and MCS-12 indicate better improvement.

Lower scores in NPRS and RMDQ indicate better improvement.

Participants in LSE_{RUSI-biofeedback} group = 30, LSE_{no-biofeedback} group = 30, and minimal intervention group = 30.

*Comparison is significant at the 0.05 level,

**Comparison is significant at the 0.01 level.

***Comparison is significant at the 0.001 level.

^†p-value for interaction effect.

Fig. 3.

Lumbar multifidus cross-sectional area scores for all groups across time. LSE, lumbar stabilization exercise; RUSI, real-time ultrasound imaging.

For pain intensity, there was significant group x time interaction effect (F[4, 174] = 63.04, p < 0.001) and main effect of group (F[2, 87] = 91.28, p < 0.001) (Table 2). Post-hoc analysis with LSD indicated that both LSE_{RUSI-biofeedback} and LSE_{no-biofeedback} groups were superior to the minimal intervention group at 8-week follow-up (− 3.56 points [95% CI, − 4.01 to − 2.98; p < 0.00; effect size = 2.81] and − 4.00 points [95% CI, − 4.51 to − 3.45; p < 0.001; effect size = 5.54], respectively) and 20-week follow-up (− 3.30 points [95% CI, − 3.81 to − 2.78; p < 0.001; effect size = − 2.65] and − 3.00 points [95% CI, − 3.51 to − 2.48; p < 0.001; effect size = 3.61], respectively). However, no significant difference was observed between the experimental groups (p > 0.05) at any time point. The main effect of time was also significant (F[2, 174] = 377.58, p < 0.001), with all groups showing a decrease in the mean pain intensity over time (p < 0.001)(Fig. 4A).

Fig. 4.

(A) Pain intensity scores for all groups across time (B) functional disability scores for all groups across time (C) physical health scores for all groups across time (D) mental health scores for all groups across time. LSE, lumbar stabilization exercise; NPRS, Numeral Pain Rating Scale; RMDQ, Rolland-Mmorris Disability Questionnaire; RUSI, real-time ultrasound imaging; PCS, physical component summary; MCS, mental component summary.

Similarly, for functional disability, there was significant group x time interaction effect (F[4, 174] = 75.55, p < 0.001) and main effect of group (F[2, 87] = 98.12, p < 0.001) (Table 2). Although both LSE_{RUSI-biofeedback} and LSE_{no-biofeedback} groups were superior to the minimal intervention group at 8-week follow-up (− 7.10 points [95% CI, − 8.14 to − 6.05; p < 0.001; effect size = 3.4] and − 7.43 points (95% CI, − 8.47 to − 6.38− 7.23; p < 0.001; effect size = 5.69], respectively) and 20-week follow-up (− 7.77 points (95% CI, − 8.81 to − 6.72; p < 0.001; effect size = − 3.31] and − 7.23 points (95% CI, − 8.27 to − 6.18; p < 0.001; effect size = − 5.27], respectively), post-hoc analysis with LSD revealed no significant difference between the experimental groups (p > 0.05). The main effect of time was also significant (F[2, 174] = 324.69, p < 0.001), with both the LSE_{RUSI-biofeedback} and LSE_{no-biofeedback} groups showing a decrease in the mean functional disability over time (p < 0.001) but not in the minimal intervention group (p > 0.05) (Fig. 4B).

Regarding the physical health component of QOL, there was a significant group x time interaction effect (F[4, 174] = 36.48, p < 0.001) and main effect of group (F[2, 87] = 71.22, p < 0.001) (Table 2). Although both LSE_{RUSI-biofeedback} and LSE_{no-biofeedback} groups were superior to the minimal intervention group at all time points (p > 0.05), the LSE_{no-biofeedback} group was superior to the LSE_{RUSI-biofeedback} group by an additional − 4.89% points (95% CI, − 7.92 to − 1.86; p = 0.002; effect size = 0.83) at 20-week follow-up, as shown by post-hoc analysis with LSD. The main effect of time was also significant (F[2, 174] = 143.13, p < 0.001), with all groups showing an increase in the mean physical health over time (p < 0.001) except for the minimal intervention group at 20-week follow-up (p > 0.05) (Fig. 4C).

Again, there was a significant group x time interaction effect (F[4, 174] = 73.45, p < 0.001) and main effect of group (F[2, 87] = 69.63, p < 0.001) for the mental health component of QOL (Table 2). Post-hoc analysis with LSD showed that the LSE_{RUSI-biofeedback} group had greater improvement compared to both LSE_{no-biofeedback} and minimal intervention groups, by an additional 3.63% points (95% CI, 0.32 to 6.94; p = 0.032; effect size = 0.61) and 23.2% points (95% CI, 19.9 to 26.5; p < 0.001; effect size = 3.23) respectively, at 8-week follow-up. This superiority was maintained at the 20-week follow-up (7.34% points [95% CI, 4.03 to 10.6; p < 0.001; effect size = 1.18] and 23.1% [95% CI, 19.8 to 26.4; p < 0.001; effect size = 2.65], respectively). Moreover, the LSE_{no-biofeedback} group was superior to the minimal intervention group at all time points. The main effect of time was also significant (F[2, 174] = 32.02, p < 0.001), with the experimental groups showing an increase in the mean mental health over time, whereas the minimal intervention group showed a decline (p < 0.001) (Fig. 4D).

Association between post-treatment changes in LM CSA and pain intensity and functional disability among the experimental groups

In the LSE_{RUSI-biofeedback} group, moderate significant correlations were found between changes in LM CSA scores and NPRS scores at both 8-week and 20-week follow-ups, while weak and moderate significant correlations were observed between changes in LM CSA scores and RMDQ scores at both 8-week and 20-week follow-ups, respectively (Table 3). However, in the LSE_{no-biofeedback} group, no significant correlations were found between changes in LM CSA scores and NPRS and RMDQ scores at all time points (p > 0.05). When both the experimental groups were combined, weak significant correlations were observed between changes in LM CSA scores and NPRS and RMDQ scores at all time points except for the changes in NPRS scores at 8-week follow-up (p > 0.05) (Table 3). Furthermore, no significant correlations were observed between the variables in the minimal intervention group (p > 0.05).

Table 3.

Correlation between post-treatment changes in LMM CSA scores and pain intensity and functional disability scores in the experimental groups.

	∆ LM CSA		∆ LM CSA
	8 weeks r (95% CI)	p-value	20 weeks r (95% CI)	p-value
LSERUSI-biofeedback (n = 30)
∆ NPRS (0–10)	− 0.40 (− 6.79 to − 0.03)	0.027*	− 0.50 (− 0.77 to − 0.20)	0.005**
∆ RMDQ (0–24)	− 0.38 (− 0.67 to − 0.01)	0.037*	− 0.42 (− 0.74 to 0.50)	0.022*
LSEno-biofeedback (n = 30)
NPRS (0–10)	− 0.66 (− 0.35 to 0.41)	0.729	− 0.08 (− 0.44 to 0.21)	0.639
RMDQ (0–24)	− 0.18 (− 0.43 to 0.16)	0.345	− 0.05 (− 0.37 to 0.27)	0.813
Combined (n = 60)
∆ NPRS (0–10)	− 0.23 (− 4.14 to − 0.13)	0.077	− 0.37 (− 0.51 to − 0.23)	0.004**
∆ RMDQ (0–10)	− 0.32 (− 0.50 to − 0.13)	0.013*	− 0.33 (− 0.51 to − 0.18)	0.009**

CI, confidence interval; LSE, lumbar stabilization exercise; BMI, body mass index; CSA, cross-sectional area; LM, lumbar multifidus; r, Pearson’s correlation coefficient.

*p < 0.05.

**p < 0.01.

Discussion

To the best of our knowledge, this is the first RCT to examine if the incorporation of RUSI biofeedback during LSE would enhance LM CSA, pain intensity, functional disability, and QOL compared to LSE without biofeedback in patients with NCLBP.

Although the LSE program effectively improved LM CSA, both with and without biofeedback, compared to minimal intervention, adding biofeedback using RUSI yielded a greater effect, with a large effect size being observed at 20-week follow-up. Additionally, the magnitude of the group differences (~ 2.7 points) was slightly above the MCID threshold values (2.0 points) that were used to power the study. The maintenance of this effect, even though within the span of 20 weeks, is crucial to the efficiency of the LM to maintain stability and control of the spine, which is believed to prevent the recurrence of LBP. Proper and lasting activation of the LM is, therefore, crucial in LBP rehabilitation. RCTs on the additive effect of LSE training with RUSI biofeedback on LM CSA are limited. In a small controlled laboratory study, no significant difference in LM thickness was found after a single LSE session with or without RUSI biofeedback⁴⁰, which is not on par with our study. However, the study applied a single treatment session, which may not have been adequate to elicit any significant change. On the other hand, a previous systematic review found RUSI feedback to be more effective than tactile and/or verbal biofeedback for improving LSE performance and retention success¹⁶, which is in agreement with our study.

The results that both pain intensity and functional disability were significantly improved in participants receiving LSE with and without biofeedback, compared to those receiving minimal intervention, support the efficacy and utility of this exercise program. Though the experimental groups were comparable, the percentage change in pain intensity and functional disability scores achieved by the LSE_{RUSI-biofeedback} group (57.7% and 67.9%, respectively) was slightly better than those achieved by the LSE_{no-biofeedback} group (52.6% and 61.9% respectively) at 20-week follow-up. Importantly, the changes from baseline in both groups exceeded the MCID threshold of 30% improvement proposed for both pain intensity and functional disability measured using the NPRS and RMDQ⁴¹. In this regard, our study aligns with the existing literature supporting the efficacy of LSE in alleviating pain and disability in patients with NCLBP^42,43.

Health-related quality of life is another important core outcome domain in non-specific LBP trials⁴⁴, with improving this outcome being among the key goals of rehabilitation. Consistent with our findings, previous studies have reported the effectiveness of LSE in enhancing various domains of QOL^45–50. However, only a few studies have reported QOL as physical and mental summary scores using the SF-12 or SF-36 health surveys^26,51. Improvement in QOL may be attributed to improvement in pain and function^52,53 as reduced QOL has been closely linked to these outcomes in NCLBP⁵⁴. Notably, we found that individuals receiving LSE_{RUSI-biofeedback} had superior improvement compared to those receiving training without biofeedback or minimal intervention only for mental health at all time points, whereas training without biofeedback was superior for physical health at 20-week follow-up. The LSE_{RUSI-biofeedback} group may have shown greater mental health gains due to the motivational impact of visual biofeedback, enhancing confidence and self-efficacy. In contrast, better physical health in the LSE_{no-biofeedback} group may reflect simpler, physical engagement with the exercises. Supporting our findings, a pilot study by Shaughnessy and Caulfield⁴⁸ reported significant improvements across various domains of QOL using the SF-36 health survey in favor of the LSE program compared to no intervention. Specifically, greater improvements were observed in physical health domains (physical functioning, role physical, and bodily pain) and mental health domains (vitality, social functioning, role emotional, and mental health)⁴⁸. Similarly, Burut and colleagues reported significant improvements in mental health, physical role, bodily pain, and vitality domains in favor of LSE plus conventional physiotherapy over conventional physiotherapy alone⁵⁵. In contrast, other studies reported comparable effects on physical and mental health between LSE and graded activity⁵⁰, as well as patient education⁵¹.

The results showed that post-intervention changes in LM CSA scores were weakly to moderately correlated with changes in pain intensity and functional disability scores. However, the study cannot verify the hypothesis that an increase in the LM size is influenced by changes in pain and disability. Instead, it appears that an increase in the LM size is related to changes in pain and disability. This is contrary to the conclusion of a previous systematic review²¹, which is based on the inclusion of two studies, reported no evidence that post-LSE increases in LM resting thickness in patients with acute non-specific LBP⁵⁶ or LM CSA in patients with NCLBP⁵⁷ were related to a reduction in pain. Additionally, no evidence suggested that post-LSE increases in LM CSA were related to changes in disability⁵⁶. It is worth noting, however, that only one study was conducted on patients with NCLBP, and the authors did not incorporate RUSI biofeedback in LSE training⁵⁷. The lack of correlation between changes in LM thickness and pain in the acute non-specific LBP study⁵⁶ could be due to the fact that reduced LM thickness, which also reflects activation deficit, is likely to be caused by neural response in acute LBP, whereas reduced LM CSA results from disuse atrophy, fatty infiltration, change in muscle fiber type, which is commonly seen in chronic LBP⁵⁸. The lack of correlation observed in the NCLBP study⁵⁷, however, may be attributed to differences in intervention protocols, such as the absence of incorporating RUSI biofeedback or variations in the LSE training methods used. Further studies are warranted to verify these associations to enhance understanding of the role of LM in LBP.

This study has some limitations, which should be taken into consideration when interpreting the results. First, LM CSA was measured only at the L5 level, limiting the detection of changes at other lumbar levels despite evidence of L5-specific aberrations and recovery with rehabilitation^25,59. Second, we could not rule out potential variability associated with ultrasound measurements of LM CSA, which can be operator-dependent and sensitive to subtle differences in probe placement and participant positioning, despite efforts to standardize the procedure. Third, the delivery of LSE by a single, highly experienced clinician may have introduced performance bias, potentially influencing the outcomes of the experimental groups. Fourth, incorporating RUSI biofeedback into the LSE program requires additional time, training, and resources for both patients and providers; hence, its use may be limited in clinics without access to ultrasound equipment. Fifth, although ITT analysis was conducted, the high attrition rate in the minimal intervention group at the 20-week follow-up may still bias the results by reducing group comparability and potentially overestimating the effectiveness of the experimental groups. Sixth, this study evaluated outcomes only in the short term, which limits our understanding of the long-term sustainability of the observed effects. It would be crucial to determine whether the changes in LM CSA are maintained over time, given the high recurrence rate of LBP. Lastly, while our study focused on the LM, excluding the TrA would limit interpretation of the intervention’s full neuromuscular impact. Future research should assess the TrA and consider incorporating diaphragm retraining, given its synergistic role in core stability alongside the LMM, pelvic floor, and TrA.

Conclusions

This study suggests that LSE training, with or without RUSI biofeedback, led to short-term improvements in LM CSA, pain intensity, functional disability, and QOL. However, LSE with RUSI biofeedback led to greater effects on LM CSA and mental health compared to training without biofeedback. Furthermore, an increase in LM CSA appears to be related to changes in pain intensity and functional disability.

Abbreviations

CSA

Cross-sectional area

ITT

Intention to treat

LBP

Low back pain

LM

Lumbar multifidus

LSD

Least significant difference

LSE

Lumbar stabilization exercises

NCLBP

Nonspecific chronic low back pain

NPRS

Numerical Pain Rating Scale

MCS-12

Mental component scores

PCS-12

Physical component scores

QOL

Quality of life

RCT

Randomized controlled trial

RMDQ

Roland Morris Disability Questionnaire

RUSI

Real-time ultrasound imaging

Author contributions

RS and SOG conceived the study. RS and AIA developed the first draft of the manuscript. MKO, BK, BB, AWA, and PW reviewed and edited the manuscript. All authors contributed to the trial design and have read, contributed to, and approved the final manuscript.

Funding

This research received no grant from any funding agency.

Data availability

After the study is completed, the data will be made available on request from the corresponding author.

Declarations

Competing interests

The authors declare no competing interests.

Ethical approval

The Health Research Ethics Committee of National Orthopedic Hospital, Dala, Kano approved this study (Ref: NOHD/RET/ETHIC/60).