A single-blinded (evaluator), parallel-groups, randomized clinical trial was prospectively registered by the number clinical trial NCT04582812 at ClinicalTrials.gov and performed from 30 November 2022 to 19 February 2024, following the Consolidated Standards of Reporting Trials (CONSORT 2010) criteria (Schulz et al., 2010).

The Helsinki Declaration and all ethical requirements regarding human experimentation were respected (Holt, 2014). The study was approved on 18 November 2020 by the ethics committee of the San Carlos Clinical Hospital (Madrid, Spain) with the approval code 20.655-E_BS. All patients included in the present study received the information sheet and signed the informed consent form.

This research study was supported and funding by the Spanish Ministry of Science, Innovation and Universities and the State Agency for investigation of the national government regarding the Call for Innovation, Development and Research (I + D + i Projects) in 2020, according to the framework of the State Program for Knowledge Generation for Scientific and Technological Strengthening of the I + D + i System as well as I + D + i oriented to Society Challenges with the grant number PID 2020-117162RA-I00 funded by MICIU/AEI/10.13039/501100011033.

A patent registry was performed for a utility model in the Spanish Patent and Trademark Office with application number U202200045, publication number ES1288519, and issue date of 30 March 2022. This bilateral thoracic orthosis, including both the right and left holding devices for two ultrasound probes, showed reliability from good to excellent and adequate repeatability for the simultaneous thickness measurement of both hemidiaphragms bilaterally during normal breathing. The use of this device was previously recommended for simultaneous breathing reeducation of both the right and left hemidiaphragms by ultrasonography visual biofeedback (Molina-Hernández et al., 2023).

The sample size calculation for this study was carried out using the difference for two independent groups by the 3.1.9.2 version of the G*Power program (G*Power^©; Dusseldorf University; Germany) (Faul et al., 2007), considering the difference in the thickness of the left hemidiaphragm during inspiration as the main outcome, given that this measurement was associated with muscular alterations in the lumbar region, as this hemidiaphragm was claimed to play a key role in postural function (Celli, 1989; Hruska, 1997; Terada et al., 2016), and a moderate effect size with a Cohen’s d of 0.63 necessary to normalize the difference in diaphragmatic thickness of patients with NSLBP with respect to healthy subjects (Calvo-Lobo et al., 2019), using a two-tailed hypothesis, a probability of error α of 0.05, a power of 0.80 (1-β probability of error), and a randomization rate of 1 (N₂/N₁). According to these parameters, a sample size of 82 patients with NSLBP was necessary to achieve an actual power of 0.804, divided into two groups of 41 patients in each intervention group. Considering a 15% possible loss to follow-up, 96 patients with NSLBP were recruited for the total sample size, including 48 patients with NSLBP per group.

A total sample of 96 patients with NSLBP was recruited by announcements from the different health sciences faculties of the Complutense University of Madrid (Madrid, Spain) from 30 November 2022 to 23 March 2023. A sex-based stratified random sampling method was developed to recruit 48 men and 48 women in order to determine the influence of sex on the outcomes that showed effectiveness after comparing both interventions. Both treatment groups were pair-matched by sex, including 24 men and 24 women for each intervention group (Berndt, 2020).

Inclusion criteria comprised patients, including students, professors, or persons coming from outside the university, with bilateral NSLBP medical diagnosis with duration for more than 6 weeks and a pain location mainly between the bi-iliac line and the subcostal line as well as a positive active straight leg raise test bilaterally, aged between 18 and 65 years (Patel et al., 2016; Calvo-Lobo et al., 2019). Exclusion criteria comprised congenital lumbar disorders, rheumatic or neuromuscular disorders, body mass index (BMI) greater than 31 kg/m², previous diagnosis of respiratory or neurological pathology, previous surgery and lower limb pathology (including fractures, sprains, or joint instability), skin disorders, pregnancy, inability to follow instructions during the study, and presence of hyperventilation syndrome as assessed by the Nijmegen test indicated by a score equal to or greater than 24 points (Paris-Alemany et al., 2018; Calvo-Lobo et al., 2019; Marugán-Rubio et al., 2021).

A simple randomization process was performed by applying based-sex stratification in order to allocate the 96 patients with NSLBP into both intervention groups (including 24 men and 24 women for each group) using the 4.1 version of the EPIDAT program (Xunta de Galicia; Conselleria de Sanidade; Galicia; Spain) (Berndt, 2020).

According to this randomization process, 48 patients (24 men and 24 women) were assigned to high-intensity inspiratory muscle self-training intervention for 8 weeks (Janssens et al., 2015; Marugán-Rubio et al., 2022) (IMT group) and 48 patients (24 men and 24 women) were also assigned to the same IMT for 8 weeks plus simultaneous bilateral visual biofeedback of the diaphragm by ultrasound imaging for 6 weeks (BFB + IMT group) (Hides et al., 2008; Janssens et al., 2015; Marugán-Rubio et al., 2022; Molina-Hernández et al., 2023). Therapists with more than 4 years of experience in IMT and BFB applied both interventions. Outcome measurements were performed at baseline and after 8 weeks of intervention by a physician together with other evaluators experienced in ultrasound imaging of the diaphragm who were blinded to the intervention group allocation using numerical coding (Marugán-Rubio et al., 2022).

The IMT group received only high-intensity inspiratory muscle self-training for 8 weeks. Patients were instructed to breathe using a mouthpiece (POWERbreathe, Medic, HaB International Ltd., Warwickshire, United Kingdom) that occluded their nose in a standing position and produced a negative pressure according to 60% of their maximum inspiratory pressure (MIP) using an inspiratory valve that resisted each inspiration effort (Figure 1A). Patients performed 30 breaths twice per day, 7 days per week, with a rate of 15 breaths per minute and applying a duty cycle of 0.5. Furthermore, all NSLBP participants were trained to apply mainly diaphragmatic breathing, named the “bucket-handle” motion, rather than thoracic breathing, named the “pump arm” motion, through the provision of tactile and verbal signals (Janssens et al., 2015; Marugán-Rubio et al., 2022).

The BFB + IMT group received the same high-intensity IMT for 8 weeks plus simultaneous bilateral visual biofeedback of the diaphragm by ultrasound imaging for 6 weeks using the proposed bilateral thoracic orthotic device for reeducation of both hemidiaphragms at the same time during normal breathing (Figure 1B). Patients were instructed in the same way for high-intensity IMT self-training in conjunction with diaphragmatic breathing reeducation using ultrasound visual biofeedback with the proposed thoracic orthosis to facilitate probe fixation and visualization of the ultrasound screen, selectively explaining diaphragmatic thickening during inspiration and correcting paradoxical breathing patterns (Hides et al., 2008; Janssens et al., 2015; Calvo-Lobo et al., 2019; Marugán-Rubio et al., 2022; Molina-Hernández et al., 2023). The same two high-quality ultrasonography tools (E-CUBE i7, Alpinion Medical Systems, Seoul, Korea) and the two linear probes (Broadband Linear type L3_12T; 38.4 mm field of view, 128 elements) with a frequency range from 8.0 to 12.0 MHz and a 45 mm probe foot were used for the measurements and the interventions. Transcostal diaphragmatic thickness was evaluated by B-mode imaging at rest in a supine position with a preset depth of 3 cm, 12 MHz frequency, 64-point gain, 64-point dynamic range, and one focus located at 2 cm depth (Calvo-Lobo et al., 2019; Marugán-Rubio et al., 2022; Molina-Hernández et al., 2023). The registered patent comprised two holding devices that fixed the right and left ultrasound probes to the thoracic orthosis by two bivalve adapters, permitting the insertion and fixation of both ultrasound probe holders without interfering with the patients’ breathing patterns (Marugán-Rubio et al., 2021; Molina-Hernández et al., 2023). This device allowed total thoracic mobility. The addition of ultrasound gel in the spaces below each probe footprint permitted a complete visualization of the last intercostal space. Both the right and left linear ultrasound probes were bilaterally placed perpendicular to the last intercostal space following the right and left mid-axillary lines of the patient, who was located in the supine decubitus (Harper et al., 2013; Molina-Hernández et al., 2023). This bilateral thoracic orthosis presented good to excellent reliability and adequate repeatability with an intraclass correlation coefficient (ICC) of 0.614 to 0.997, a standard error of measurement (SEM) of 0.002 to 0.028 cm, and minimum detectable change (MCD) of 0.006 to 0.079 cm for the bilateral simultaneous thickness evaluation of both hemidiaphragms during normal breathing. The orthosis was previously recommended for simultaneous breathing reeducation of both the right and left hemidiaphragms by ultrasonography visual biofeedback (Molina-Hernández et al., 2023).

Descriptive data such as sex (dichotomous categorical variable, male or female), age (continuous quantitative variable detailed in years), height (continuous quantitative variable detailed in cm), weight (continuous quantitative variable described in kg), and BMI (continuous quantitative variable detailed in kg/cm² according to the Quetelet index (Garrow, 1986)) were detailed (Hides et al., 2008; Janssens et al., 2015; Calvo-Lobo et al., 2019; Marugán-Rubio et al., 2022; Molina-Hernández et al., 2023). The International Physical Activity Questionnaire (IPAQ), a questionnaire with adequate psychometric properties, was applied to determine the rate of metabolic equivalents per task per minute per week (METs/min/wk; measured as a continuous quantitative variable) and categorized according to the level of physical activity into level I—sedentary (<600 METs/min/wk), level II—moderate (600–1500 METs/min/wk), or level III—vigorous (>1500 METs/min/wk), as a polytomous categorical variable (Gauthier et al., 2009; Marugán-Rubio et al., 2022; Molina-Hernández et al., 2023).

For the main and secondary purposes of this research study, the Statistical Package of Social Sciences (SPSS) version 24.0 software application (IBM; Armonk, NY, IBM Corp.) was used for the data analysis to compare descriptive data and primary and secondary outcomes between both intervention groups. An α error of 0.05 was applied, and p-values <0.05 were determined to be statistically significant for a confidence interval (CI) of 95%. Raw data of the present study may be accessed at Data File S1. All analyses were carried out regarding two intervention groups and the difference between two measurement times (baseline before intervention and after 8 weeks of intervention). The Kolmogorov–Smirnov statistical test was applied to determine the distribution normality. This statistical test was recommended in the field of health sciences for sample sizes larger than 30 patients per group (Ghasemi and Zahediasl, 2012). Mean ± standard deviation (SD), lower and upper limits according to 95% CI, and range (minimum and maximum values) were determined for all data. Student’s t-tests were used to compare differences between the two independent groups regarding parametric data using the p-value following Levene’s tests of variance equality (p-value <0.05 if there were no variance equalities). The differences between the two independent groups regarding non-parametric data were determined by Mann–Whitney U tests. Frequencies (n) and percentages (%) were applied to describe all categorical data. The comparison of categorical data among groups was performed by Fisher’s exact tests if dichotomous variables were analyzed or chi-squared (χ²) tests if polytomous variables were tested. Moreover, the effect sizes of the comparisons between the two intervention groups for the primary and secondary outcomes were calculated by applying Cohen’s d calculated by the formula d = 2 t / g d l , considering the categorization of very small effect size if d was lower than 0.20, small effect size if d varied from 0.20 to 0.49, medium effect size if d ranged from 0.50 to 0.79, and large effect size if d was equal or higher than 0.80 (Cohen, 1988; Calvo-Lobo et al., 2019; Marugán-Rubio et al., 2022).

Following the last aim of this research study, the influence of age and sex on the outcomes that showed effectiveness under this treatment in NSLBP patients was determined by analyses of covariance (ANCOVA). Multivariate linear regression analyses were performed to determine the influence and prediction of the statistically significant outcome measurement differences between both interventions according to the analyses described above (i.e., left diaphragm thickness at T^ins and T^ins-exp, right and left PPTs) based on the intervention group, sex, and age. First, ANCOVAs for repeated measures with linear graphs were performed for each statistically significant outcome considering two groups (IMT and BFB + IMT groups) × two times (baseline and after intervention measurement moments) × covariables (age and sex) considering the p-value and F statistic according to the Greenhouse–Geisser correction if the Mauchly test rejected the sphericity completed with the partial eta squared coefficient (η_p ²), considering η_p ² = 0.01 as a small effect size, η_p ² = 0.06 as a medium effect size, and η_p ² = 0.14 as a large effect size (Cohen, 1973; Blanca et al., 2017). Second, each linear regression analysis was performed for each statistically significant outcome using the “stepwise selection” method. Each regression coefficient (R ²) was calculated to determine the adjustment quality (Austin and Steyerberg, 2015). In addition, age (years), sex (female = 1; male = 2), and intervention group (IMT = 1; BFB + IMT = 2) were considered as independent variables for each linear regression analysis. Each outcome measurement that presented statistically significant differences between the intervention groups was included for each linear regression analysis as a dependent variable. The pre-set F probabilities were P_in and P_out of 0.05 and 0.10, respectively (Calvo-Lobo et al., 2019; Marugán-Rubio et al., 2022).

Of 120 patients with NSLBP assessed for eligibility, 24 participants were excluded secondary to lumbar surgery (n = 2), not meeting NSLBP criteria of the study (n = 5), NSLBP shorter than 6 weeks (n = 2), age over 65 years (n = 2), IMC greater than 31 kg/cm² (n = 4), presence of other pathologies (n = 4), and refusal to participate (n = 5). Thus, a total of 96 patients with NSLBP were randomized into BFB + IMT (n = 48) and IMT (n = 48). In each group, three patients did not receive the allocated intervention due to not attending at the baseline time for the initial evaluation, and the remaining patients were assessed at baseline and received the allocated intervention in each group (n = 45). During the follow-up, three participants were lost in each group due to not attending follow-up meetings, work leave, and IMT side effects such as abdominal discomfort and dental problems during IMT. Finally, 84 patients with NSLBP were analyzed (n = 42 patients in each group), according to Figure 3.

The baseline sample was pair-matched by sex (p = 1.00) and comprised 90 patients with NSLBP divided into a BFB + IMT group (n = 45), which included 23 (51.11%) female and 22 (48.89%) male patients, and an IMT group (n = 45), which comprised 22 (48.89%) female and 23 (51.11%) male patients. The total sample showed a mean ± SD of age of 47.66 ± 11.10 years and normal BMI of 25.10 ± 3.16 kg/cm², being homogeneous (p > 0.05) regarding the age, height, weight, BMI, and IPAQ scores between both groups. Furthermore, the IPAQ physical activity levels were similar (p = 0.391) between both groups because the BFB + IMT group included six sedentary patients, 14 participants with moderate activity, and 25 patients who performed vigorous physical activity, while the IMT group comprised six sedentary patents, 20 patients with moderate activity, and 19 patients who performed vigorous physical activity. Table 1 shows the baseline quantitative descriptive data of the BFB + IMT and IMT groups.

Although the total sample (n = 96) was randomized, only 90 patients were assessed at baseline due to the non-presence of six patients for the initial evaluation. This sample was homogenous for all respiratory outcomes at baseline because there were not statistically significant differences (p > 0.05) between both intervention groups for bilateral diaphragm thickness during normal breathing, respiratory muscle strength, and lung function spirometry parameters. These findings are presented in Table 2.

The sample was also homogenous for all clinical outcomes at baseline because there were no significant differences (p > 0.05) between both groups for pain intensity, bilateral PPT of paraspinal muscles, disability, or quality of life for total scores and physical and mental health domains. These findings are presented in Table 3.

After 8 weeks of intervention, the BFB + IMT group showed statistically significant differences (p < 0.05) with an increase in the left hemidiaphragm thickness difference at T^ins with a medium effect size (d = 0.53) and T^ins-exp with a small effect size (d = 0.38) compared to the IMT group. The remaining diaphragm thickness differences after 8 weeks of both interventions did not show statistically significant differences (p > 0.05) with an effect size from very small to small (d = 0.14–0.38). These results are shown in Table 4.

There were no statistically significant differences (p > 0.05) with an effect size from very small to small (d = 0.10–0.23) for the other respiratory secondary outcomes, including respiratory muscle strength and lung function spirometry parameters. These findings are shown in Table 5.

After 8 weeks of intervention, the BFB + IMT group presented statistically significant differences (p < 0.01) with an increased right and left PPT with a medium effect size (d = 0.71–0.74) with respect to the IMT group. The remaining clinical outcome differences did not show statistically significant differences (p > 0.05), with an effect size from very small to small (d = 0.02–0.46). These findings are presented in Table 6.

The age and sex influence on the outcome differences (i.e., left diaphragm thickness at T^ins and T^ins-exp, right and left PPTs) that showed statistically significant differences after 8 weeks of BFB + IMT versus IMT in NSLBP patients was analyzed by ANCOVAs for repeated-measures according to the Greenhouse–Geisser correction and predicted by multivariate linear regression analyses based on age, sex, and group as independent variables.

First, the interactions of time × group (p = 0.003; F_(1,81) = 9.739; η_p ² = 0.107) and time × sex (p = 0.007; F_(1,81) = 7.756; η_p ² = 0.087) were statistically significant with medium effect sizes for the left hemidiaphragm thickness difference at T^ins (Figure 4A). However, the interaction of age × group (p = 0.233; F_(1,81) = 1.442; η_p ² = 0.018) was not statistically significant with a small effect size. A linear regression model (R ² = 0.178) predicted a higher left hemidiaphragm thickness difference at T^ins based on the BFB + IMT group (R ² = 0.099; β = 0.050; F_(1,82) = 8.997; p = 0.004) and male sex (R ² = 0.079; β = 0.045; F_(1,81) = 7.756; p = 0.007).

Second, the interaction of time × group (p = 0.045; F_(1,81) = 4.130; η_p ² = 0.049) was also statistically significant with a small effect size, but neither time × sex (p = 0.447; F_(1,81) = 0.585; η_p ² = 0.007) or time × age (p = 0.084; F_(1,81) = 3.052; η_p ² = 0.037) with small effect sizes was statistically significant for the left hemidiaphragm thickness difference at T^ins-exp (Figure 4B). A linear regression model (R ² = 0.052) predicted a greater left hemidiaphragm thickness difference at T^ins-exp based on younger age (R ² = 0.052; β = −0.001; F_(1,82) = 4.540; p = 0.036).

Third, time × group interaction (p = 0.001; F_(1,81) = 11.501; η_p ² = 0.124) was also significant with a medium effect size, although neither time × sex interaction (p = 0.799; F_(1,81) = 0.065; η_p ² = 0.001) nor time × age interaction (p = 0.755; F_(1,81) = 0.098; η_p ² = 0.001) with small effect sizes, was effective for the right PPT of the paraspinal muscles (Figure 4C). The multivariate analysis did not display a valid linear regression model due to a non-significant constant p-value according to pre-set F probability considering P_in and P_out of 0.05 and 0.10, respectively.

Lastly, the interaction of time × group (p = 0.002; F_(1,81) = 10.587; η_p ² = 0.116) also presented significant differences with a medium effect size, but neither time × sex interaction (p = 0.952; F_(1,81) = 0.004; η_p ² = 0.000) nor time × age interaction (p = 0.292; F_(1,81) = 1.124; η_p ² = 0.014) with small effect sizes, were effective for the left PPT of the paraspinal muscles (Figure 4D). In line with the last analysis, the multivariate analysis did not display any valid linear regression model following the non-significant constant p-value of the pre-established F probability values for P_in of 0.05 and P_out of 0.10.

Further studies should control the effectiveness of the proposed simultaneous bilateral reeducation of the diaphragm, including a third arm with a sham IMT intervention (Janssens et al., 2015). In addition, other musculoskeletal conditions could benefit from the simultaneous reeducation of both hemidiaphragms. For example, women with fibromyalgia showed positive effects in respiratory efficiency and quality of life after IMT, and the proposed simultaneous bilateral diaphragmatic reeducation could improve these results (Tomas-Carus et al., 2022).

The lack of a control group with a sham IMT may be considered the main limitation of our study (Vicente-Campos et al., 2021). High-intensity IMT was proposed in both groups because this intervention was shown to be more effective than low-intensity IMT in patients with NSLBP, and other IMT intensities should be considered in future studies (Janssens et al., 2015). The authors acknowledge that the specificity of the inclusion and exclusion criteria, which could potentially restrict the generalizability of the study’s findings, could be a limitation. Specifically, criteria such as an age limitation of 65 years and older and a BMI greater than 31 kg/cm² may be noteworthy because it is not uncommon for older and obese patients to experience NSLBP (Maher et al., 2017).