This randomized controlled trial was conducted at the Universidade Federal de Pelotas (UFPel) and the Sindicato dos Policiais Rodoviários Federais do Rio Grande do Sul, Brazil (Fig. 1). The study was approved by the Research Ethics Committee of the Escola Superior de Educação Física at UFPel (protocol 58,715,422.3.0000.5313), and followed the CONSORT guidelines.18,19 All individuals provided signed informed consent before participation. The trial was prospectively registered on ClinicalTrials.gov (NCT05481996), and the full protocol is available online.20

Fig. 1.

Recruitment of participants, exclusions, and randomization of participants with chronic low back pain included in the study and follow-ups.

We enrolled actively working PSW (police officers and firefighters working in public security organizations in Rio Grande do Sul state), aged 18–60, with chronic LBP (Numerical Rating Scale score ≥3/10).21 All participants were engaged in either operational or administrative duties at the time of enrollment. Participants had to have an Android smartphone and email access. Chronic LBP was defined as discomfort between the 12th ribs and lower gluteal folds, lasting for >12 weeks, without a specific diagnosis.22 Exclusions included neurological symptoms, severe spinal diseases, cardiovascular/metabolic conditions, recent spine surgery (over the last 12 months), pregnancy, recently receiving physical therapy or performing exercise (strength training for core muscles, Pilates, yoga) within the last three months, and exercise contraindications. The Portuguese version of the Physical Activity Readiness Questionnaire (PAR-Q) was used to pre-screen for physical activity participation at baseline to ensure suitability and safety for engaging in the exercise program.23 Participants were recruited via social media advertising, local newspapers, and professional email from public security organizations from October 2022 to May 2023.

The lead researcher conducted video call interviews to assess eligibility criteria. Eligible participants received consent and standardized forms via email. A second video call with a blinded assessor collected baseline data before randomization. In-person neuromuscular assessments followed. Participants received a yoga mat and elastic band post-assessment. Assessors were uniformly trained to administer questionnaires and neuromuscular tests.

Primary outcomes were pain intensity and disability measured post intervention (i.e., at 8 weeks post-randomization). The NRS quantified average pain intensity over the last 7 days, ranging from 0 (no pain) to 10 (worst pain).21 The Roland Morris Disability Questionnaire assessed LBP-related disability using a 24-item functional assessment, with higher scores indicating more significant disability.24,25

Pain intensity and disability were assessed as secondary outcomes at 16 weeks post-randomization. Pain-related quality of life was evaluated using the WHOQoL-Pain,26 while self-efficacy was measured with the Chronic Pain Self-Efficacy Scale.27 Higher scores on both instruments indicate better quality of life and self-efficacy. Depression, anxiety, and stress symptoms were assessed using the Depression Anxiety Stress Scale,28 and sleep quality was measured with the Pittsburgh Sleep Quality Index.29 Higher scores on these scales indicate more severe symptoms. Work ability was evaluated with a single-item question regarding any change in job demands due to back and/or leg pain.30,31

Isometric trunk flexor and extensor endurance were measured through a test of sustained isometric contraction, where longer durations indicate better endurance.32–34 The maximum isometric strength of trunk extensors and flexors (Maximum Voluntary Isometric Contraction—MVIC) was assessed using a calibrated load cell (Miotec®, Porto Alegre, Brazil) of 200 kgf. Muscular activation of spine flexors (rectus abdominis) and extensors (erector spinae-longissimus) was measured simultaneously to the MVIC using surface electromyography (sEMG). The protocol description and neuromuscular assessment illustration are available in the Supplementary material.

During the initial phase (September to December 2022), eligible participants were randomly assigned to either the intervention (INT) or control (CON) group in a 1:1 ratio. However, this initial simple randomization resulted in uneven group sizes, leading to a second randomization phase (March to August 2023) to balance participant numbers across treatment arms. The second randomization involved two initial blocks with a size of 8 and a ratio of 2:1 (control: intervention) to address group imbalances, followed by 6 blocks of size 4 and a final block of size 6, all with a 1:1 ratio. Both randomization phases utilized computer-generated methods. Allocation sequences were arranged by an independent investigator and concealed in sequentially numbered, sealed, opaque envelopes. Following randomization, participants received login credentials via email to access the designated application (app) version.

Given the nature of the intervention, participants and treatment providers could not be blinded. However, outcome assessors remained blinded to treatment allocation. Participants were instructed not to disclose their assigned groups or discuss interventions during outcome assessments, and no instances of unblinding occurred.

We adhered to the TIDieR (Template for Intervention Description and Replication) checklist guidelines for telehealth interventions in clinical trials.35 The interventions used a smartphone web app for eight weeks ("MY SAFE BACK"), developed by two Computer Science Ph.D. candidates under the supervision of a professor with a decade of software development experience. The app was developed using the Adonis framework, integrated with Firebase, and utilized MySQL for the database. Participants were instructed to self-manage the functions available in the app interfaces.

Participants were allocated to one of two app versions: m-health-based core training exercise and health education (INT); or m-health-based health education (CON). Before the intervention began, participants received a message and a video tutorial on how to install and self-manage the functions available in the app (https://mysafebackapp.web.app/dashboard).

The "MY SAFE BACK" app interface for the INT group featured four sections on the main screen: 1) an online booklet, 2) weekly messages, 3) contact support, and 4) an eight-week physical exercise program (Supplementary material - Fig. 2).

The online booklet contained general information on self-managing chronic LBP. Weekly messages included content from the booklet, offering insights on exercise benefits, motivation, and positive coping strategies for pain management (Supplementary material - Fig. 3). Additionally, participants had direct access to contact the researcher for app-related queries or to report any technical issues.

The exercise component (app exercises function) comprised 16 training sessions spread across eight weeks, with participants undertaking two sessions per week on non-consecutive days (training program provided in Supplementary material - Fig. 4). Developed by professionals with over five years of clinical expertise in chronic pain management through exercise, the program targeted static core muscle endurance and dynamic core stability improvement. We structured the regimen with a progressive training approach, adhering to established protocols for chronic LBP treatment (Supplementary material - Table 1 for training progression details).36–39

The training sessions were divided into three segments: warm-up (5 min), main session (20–25 min), and cooldown (5 min) (Fig. 2– Panel A). The main portion was conducted in a circuit format, with a consistent 30-second rest interval between exercises and an identical sequence in every session. A comprehensive list featuring illustrations of exercises performed at each stage is available in Supplementary material - Table 2. Following each exercise session, participants were prompted to assess their overall perceived exertion using the Borg scale (ranging from 0 to 10 points) displayed on the app.40

Fig. 2.

Exercise panel of the My Safe Back. Training session (panel A), and Graphics Interchange Format (GIFs), descriptions, and audio on how to perform each exercise (panel B and C).

The app offered illustrated instructions using animated Graphics Interchange Format and descriptions and audio guidance for executing each exercise (Fig. 2– Panels B and C). Repetition counts or time under tension for each exercise were provided with target ranges (e.g., 10 to 12 repetitions, 20 to 30 s). Participants were prompted to indicate within the app if they could not meet these targets. All exercise sessions were conducted remotely and self-directed by the participants, with each week's sessions accessible through the app.

The app provided a count and progression bar ( %) of the number of completed training days to improve study adherence (Supplementary material - Fig. 4 - Panel A). Participants also received a text message in week 4 to encourage them to keep doing the exercises and ask questions whenever they had.

Adverse events were systematically monitored through the app starting from the second week of the intervention. At the beginning of each week, participants were prompted within the app to answer two questions: "Did you experience any new symptoms or worsening of existing symptoms this past week?" If they responded "yes," a follow-up question was displayed: "Do you think the treatment caused this? (yes/no)." Reports of adverse events were immediately reviewed by two independent researchers, and, when necessary, medical experts were consulted to determine the potential relationship to the intervention and appropriate measures to address the issue. Participants could have a video call with a physical therapist to assess issues or consider stopping treatment if it was believed to cause harm.

"MY SAFE BACK" main screen for those in the CON group included three subunits: 1) an online booklet with self-management information on chronic LBP, 2) weekly messages, and 3) contact. The contents of the booklet and weekly messages were identical to those provided to the INT group. They also received a text message in week 4 to encourage them to use the app.

Initially, we aimed for a sample size of 66 participants, expecting a significant difference of 1.5 pain intensity and 4 points in disability between groups. However, due to time constraints, we recruited only 47 participants before the deadline. Additionally, the specific nature of PSW, who often work in shifts, posed significant challenges. These irregular schedules limited their availability and willingness to participate, complicating recruitment further. We conducted a post hoc power analysis, which indicated a statistical power of approximately 63 % based on the observed effect size (0.68) and a significance level of 5 %. Though this power is lower than planned, it reflects the challenges of recruiting within our specific population of interest.

The MVIC and sEMG signals were collected and synchronized through a program implemented on the MioGraph platform (Miotec, Brazil). The full signal processing is described in the Supplementary Material (Supplementary material - Fig. 5).

We employed descriptive statistics to outline participants' baseline characteristics. Between-group differences and 95 % confidence intervals (95 % CI) for post-treatment outcomes at eight and 16 weeks were assessed using repeated-measures linear mixed models, considering participants and time as random factors. Analyses adhered to intention-to-treat principles, with missing data addressed through multiple imputation techniques. Statistical significance was set at two-sided p-values <0.05. All statistical procedures were conducted using SPSS version 20.0 for Windows (IBM corporation, Somers, New York, USA).

Those in the INT group had a greater decrease in pain intensity (MD = −1.54; 95 % CI: −2.95, −0.13) and disability (MD = −3.23; 95 % CI: −5.51, −0.95) than those in the CON group at eight weeks (Table 2). At the 16-week follow-up, those in the INT group had a greater decrease in disability (MD = −2.52; 95 % CI: −4.85, −0.20) than those in the CON group. There was no difference between groups in the decrease in pain intensity at 16 weeks (MD = −0.28; 95 % CI: −1.72, 1.16).

At 8 weeks, participants in the INT group had greater improvements than those in the CON group in total score of quality of life (MD = 11.37; 95 % CI: 3.98, 18.76), pain relief (MD = 17.60; 95 % CI: 8.37, 26.84) and uncertainty (MD = 16.46; 95 % CI: 3.31, 29.61). No between-group differences were observed in other domains at 8 weeks. At 16 weeks, the INT group showed greater improvement than the CON group only in the pain relief domain (MD =11.77; 95 % CI: 2.37 to 21.18), with no differences in change between groups in the remaining domains (Table 2).

At 8 weeks, the INT group demonstrated greater improvement in total self-efficacy (MD = 23.85; 95 % CI: 6.74, 40.96), pain management (MD = 9.70; 95 % CI: 0.69, 18.72), and coping with symptoms (MD = 13.95; 95 % CI: 7.05, 20.86) compared to the CON group. No difference between groups was observed in physical function-related self-efficacy at this time point. At 16 weeks, there were no significant differences between groups in the change in any self-efficacy domains (Table 2).

At 8 weeks, the INT group showed greater reductions in anxiety symptoms than the CON group (MD = −2.66; 95 % CI: −5.10, −0.21). This difference persisted at 16 weeks (MD = −3.10; 95 % CI: −5.59, −0.610). No between-group differences were observed for changes in depression or stress symptoms at either 8 or 16 weeks (Table 2).

There were no significant differences between the INT and CON groups in the change in sleep quality at either 8 weeks (MD = −0.46; 95 % CI: −2.02, 1.08) or 16 weeks (MD = 0.59; 95 % CI: −0.99, 2.17) (Table 2).

Regarding work ability, a consistent reduction in the percentages of work disability was observed for the participants in the INT group throughout the study. At baseline, 17.4 % of participants reported working in less physically demanding roles due to low back and/or leg pain, but this number decreased to 14.3 % after 8 weeks and to 10.5 % after 16 weeks of intervention. In comparison, the CON group, at baseline, had 12.5 % of participants reporting being in less demanding roles due to pain, and this number decreased to 8.7 % after 8 weeks, remaining stable at 8.7 % after 16 weeks.