Low back pain (LBP) is the world’s leading cause of years lived with disability. Digital exercise-based interventions have shown great potential in the management of musculoskeletal conditions, promoting access and easing the economic burden. However, evidence of their effectiveness for chronic LBP (CLBP) management compared to in-person physiotherapy has yet to be unequivocally established. This randomized controlled trial (RCT) aims to compare the clinical outcomes of patients with CLBP following a digital intervention versus evidence-based in-person physiotherapy. Our results demonstrate that patient satisfaction and adherence were high and similar between groups, although a significantly lower dropout rate is observed in the digital group (11/70, 15.7% versus 24/70, 34.3% in the conventional group; P = 0.019). Both groups experience significant improvements in disability (primary outcome), with no differences between groups in change from baseline (median difference: −0.55, 95% CI: −2.42 to 5.81, P = 0.412) or program-end scores (−1.05, 95% CI: −4.14 to 6.37; P = 0.671). Likewise, no significant differences between groups are found for secondary outcomes (namely pain, anxiety, depression, and overall productivity impairment). This RCT demonstrates that a remote digital intervention for CLBP can promote the same levels of recovery as evidence-based in-person physiotherapy, being a potential avenue to ease the burden of CLBP.
Low back pain (LBP) is a major health problem1, considered the leading cause of years lived with disability1 and of absenteeism. Although the impact on productivity varies in the literature2, one systematic review estimated the direct medical costs as $300 billion in the United States (U.S.) alone3. Unsatisfactory LBP management may lead to overutilization of imaging4, surgeries5,6, and medication, including opioids7,8.
Current guidelines for chronic LBP (CLBP) management recommend physiotherapy as a first-line intervention, alongside education and behavioral interventions9,10. Moderate-certainty evidence from randomized controlled trials (RCTs) supports the effectiveness of exercise-based physiotherapy in reducing pain and disability in LBP treatment11, and these interventions have often yielded better outcomes for disability and return to work than surgical interventions2,12.
However, access to in-person physiotherapy faces several barriers: a scarcity of healthcare resources (including therapists and facilities), time-, travel-, and costs-constraints (work time off, childcare costs), insufficient health literacy, and, more recently, the perceived risk of contracting infections13. All these also affect engagement, resulting in high percentages of unattended or incomplete treatments14.
Digital interventions have great potential in overcoming such challenges, being more accessible and affordable than in-person physiotherapy15,16, and increasing patient adherence and empowerment17. Within LBP management, research has focused on the effectiveness and safety of digital interventions, both as adjuncts to in-person care18,19 and as stand-alone through video conference-based19 or asynchronous telerehabilitation20,21. The latter has the potential to scale care delivery, addressing the growing prevalence of CLBP1. However, the few trials comparing exercise-based asynchronous interventions with standard in-person physiotherapy considered cohorts with diverse acuity levels20,21 or were non-randomized studies20, compromising the certainty of evidence on the subject. Thus, further research is needed on the effectiveness of these solutions as an alternative to in-person physiotherapy for CLBP.
Previously, we demonstrated the effectiveness of tailored digital care programs (DCP) integrating exercise, education, and cognitive behavioral therapy (CBT) in several musculoskeletal conditions22,23, including acute and chronic LBP24,25. The present RCT aims to compare the clinical outcomes of patients with CLBP following a DCP versus conventional in-person physiotherapy. We hypothesize that outcomes are comparable to those obtained with conventional physiotherapy.
Eligibility screening was conducted for 173 participants: 22 declined participation, 3 were not eligible, and 8 were excluded. In total, 140 participants were randomly assigned to the digital group (DG) or conventional group (CG) (N = 70, each). The completion rate was 81.4% (57/70) in the DG and 64.3% (45/70) in the CG (Fig. 1).
Baseline demographics were similar between groups within intention-to-treat (ITT) analysis (N = 140, Table 1). In per-protocol analysis (N = 102), no differences were observed except for age, with younger participants in the DG (Supplementary Table 1). Comparing completers (N = 102) with non-completers (N = 38), participants with lower levels of education, smokers, and those who reported not exercising at baseline were less likely to complete the program (Supplementary Table 2).
Treatment dosage was similar between groups, reflected by the similar time dedicated to exercise sessions (DG: 451.78, IQR 227.36 and CG: 385.98, IQR 145.36; P = 0.662, Table 2). On average, participants performed 22.32 (SD 9.46) exercise sessions in the DG and 12.42 (SD 4.95) in the CG. The dropout rate was higher in the CG (24/70, 34.3% vs. 11/70, 15.7%; P = 0.019).
Education and self-care tools were delivered differently to each group. The CG had support during in-person sessions, but no data was collected regarding the educational component. The DG read a median of 4.0 (IQR 6.0) educational articles and engaged with 6.0 (IQR 7.3) cognitive behavioral therapy (CBT) content pieces. Additionally, these participants communicated with the PT through a mean of 2.8 (SD 3.0) video and phone calls and 19.0 (SD 11.8) text conversations across the program. Satisfaction with the program was high and similar between groups (P = 0.837): DG: 8.4/10 (SD 2.0) and CG: 8.4/10 (SD 2.6).
Both intervention groups started with similar baseline Oswestry Disability Index (ODI) scores (median difference: −0.50, 95%CI −4.91;3.91; P = 0.821) and reported significant within-group improvements at program-end (DG: −6.90, 95% CI −19.33 to −1.57 and CG: −6.35, 95% CI: −10.36 to −3.83, both P < .001; Table 3). Changes were not significantly different between groups (median difference: −0.55, 95% CI: −2.42 to 5.81, P = 0.412), corresponding to an effect size of −0.13. The 8-week-end scores also did not significantly differ between groups (−1.05, 95% CI: −4.14 to 6.37; P = 0.671, Table 3). Per-protocol analysis yielded similar results (Supplementary Table 3).
The proportion of responders at the program end was similar between groups (DG: 40.4% (23/57) and CG: 42.2% (19/45)), denoted by a non-significant odds ratio (OR) of meeting minimal clinically important difference (MCID) (OR: 0.926, 95% CI: 0.42–2.05, P = 0.849).
Participants from both groups reported moderate pain levels at baseline (Table 3), which improved significantly in both groups (DG: −1.81, 95% CI: −3.51 to −0.25 and CG: −2.11, 95% CI −2.82 to −1.49; both P < 0.001). There were no differences in pain reduction between groups at the 8-week primary endpoint (0.21, 95% CI: −0.76 to 0.84; P = 0.913) or overall change (0.30, 95% CI −0.71 to 1.10; P = 0.666), as reflected by an effect size of −0.08.
Willingness to pursue surgery was low at baseline (DG: 6.43, 95% CI 0.00–20.68; CG: 9.76, 95% CI: 4.88–14.64), with only 34.3% (24/70) of participants in the DG and 42.9% (30/70) of participants in the CG reporting surgery intent levels above zero. Nevertheless, significant improvements were observed, with both groups reporting scores close to zero at the program-end. Both end scores and overall change were not significantly different between groups (Table 3).
Among those participants taking analgesics at baseline, 9/34 participants in the DG (26.5%) and 7/43 in the CG (16.3%) reported taking opioids (P = 0.274). Analgesic consumption remained stable in both groups (DG: P = 0.515, CG P = 0.076) until the program-end. The OR for analgesic or opioid consumption was not statistically different between groups (analgesics OR: 0.92, 95% CI: 0.00–1.23, P = 0.081; opioids OR: 0.26, 95% CI: 0.00–1.71, P = 0.985).
A similar proportion of patients reported having at least mild anxiety (Generalized Anxiety Disorder 7-item scale—GAD-7 scores ≥5: DG: 24/70 and CG:27/70) or depression (Patient Health 9-item Questionnaire—PHQ-9 scores ≥5: DG: 29/70 and CG:32/70) symptoms at baseline. Both groups showed improvements in anxiety and depression, which were not significantly different (overall change median difference: GAD-7: 0.43, 95% CI: −0.59 to 1.59; P = 0.360; PHQ-9: 1.33, 95% CI −2.20 to 2.46; P = 0.095). Regarding fear-avoidance beliefs (FAB), both groups evolved similarly as depicted by the similar 8-week end scores and changes.
Regarding physical activity levels, only the DG experienced a statistically significant within-group change towards a higher physical activity category at 8 weeks (DG: 0.66, z-score −2.71; P = 0.008 vs. CG: 0.38, z-score −1.43; P = 0.155) with no between-groups differences observed (P = 0.886, Table 4).
Very low levels of absenteeism were observed in both groups, while presenteeism, overall productivity, and non-work-related activities impairments were more expressive at baseline. Significant improvements were observed in work productivity and activity impairment (WPAI) overall, WPAI work, and non-work-related activity impairment in both groups. No differences were observed in recovery or 8-week-end scores between groups (Table 3). Absenteeism levels remained stable in both groups.
Subgroup analysis—lumbar radicular pain
Both groups comprised a similar proportion of patients with radicular LBP at baseline (DG: 21/70 (30.0%) and CG: 23/70 (32.9%), P = 0.716; Table 1).
Comparing those presenting with and without radicular pain, no differences were found in any baseline clinical score, considering both the entire cohort and individually in each group (all P > 0.05, data not shown). At the program-end, no differences were found in 8-week scores for any clinical outcome (all P > 0.05, data not shown).
Safety and adverse events
No differences were observed in adverse event rates between groups (DG: 17.5% versus CG: 10.5%, P = 0.277). No serious adverse events or hospitalizations were reported, and none of the adverse events were related to the programs (Supplementary Table 4).
This study provides evidence of the effectiveness of a digital asynchronous intervention compared to high-intensity in-person physiotherapy for CLBP. High program adherence was found in both groups, with similar treatment dosages. Dropout rates were lower in the DG. Both groups achieved significant improvements in disability and pain with no statistical differences in changes or 8-week-end scores. The proportion of treatment responders was also similar between groups (DG: 40.4% and CG: 42.2%; P = 0.849). Additionally, no significant differences between groups were found for secondary outcomes.
The comparable improvements between both rehabilitation modes are of particular importance since they illustrate the effectiveness of digital care. The digital format favors the democratization of healthcare access by overcoming geographical barriers (reaching rural or underserved areas), decreasing treatment start waiting time, and promoting patient engagement while ensuring the quality of care and potentially lowering costs. These results demonstrate that a fully-remote digital intervention for CLBP can promote similar improvements as high-intensity evidence-based in-person physiotherapy.
We observed a high enrollment, with only 12.7% of all screened individuals declining to participate, and a low dropout rate (DG: 15.7% and CG: 34.3%). This enhances generalization and supports the increased acceptance of remote care, in which the COVID-19 pandemic played a role. Similar high satisfaction levels were reported by patients from both groups, further reinforcing participants' acceptance, consistent with previous findings17.
Compliance with physiotherapy is known to be challenging13, even in the setting of RCTs26. Digital interventions have been reported to have similar16,20,27 or lower21 dropout rates than in-person physiotherapy. In the present study, the DG had lower dropout rates compared to the CG, potentially explained by the DCP convenience, reducing travel and treatment time hurdles and allowing greater flexibility on patient’s schedules, as previously reported17,28. Additionally, the context during the COVID-19 pandemic may have impacted receptivity, compliance, and overall perceptions regarding digital programs.
High engagement was observed in both groups, which translated into similar treatment dosages, as intended through the study design. The drivers for engagement with in-person physiotherapy have long been the focus of research29. In the DG, the high engagement noted might be the result of remote monitoring capabilities, which have been reported to improve self-efficacy and motivation17,28, and through bi-directional communication between patients and doctors of physiotherapy (DPTs) (including video and phone calls and asynchronously through messages)30. Frequent communication is a critical component in establishing a therapeutic alliance, a well-known factor associated with better outcomes31. The interest denoted by participants in the DG in continuing the program after the study end (data not shown) highlights the convenience of this care modality.
The specific optimal dosage of exercise (i.e., frequency and time) in CLBP rehabilitation is still unknown; therefore, recommendations range widely and are ideally personalized (2–5 times per week; 20–60 min)32. The treatment dosage reached by participants in this study falls within the recommended range.
Both groups improved similarly in disability and pain, denoted by small effect sizes (ODI: −0.13; Pain: −0.08). These results are consistent with previous trials comparing exercise-based telerehabilitation to in-person rehabilitation for patients with LBP with diverse acuity16,20,21,27, which found similar outcomes in disability and pain. These results indicate that digital interventions yield similar results to those achieved with in-person care, even in chronic conditions.
Co-prevalence of pain and mental distress is very common among chronic musculoskeletal conditions33 due to a strong and complex bidirectional relationship34, contributing to a poorer prognosis35. Exercise-based telerehabilitation interventions can positively impact mental health across several conditions24,36, including CLBP25,37. In this study, both groups experienced significant improvements in anxiety and depression, with no significant differences between groups. Likewise, FABs have been associated with non-recovery38. Herein, both groups started with comparable baseline scores and reported similar improvements, supporting that physiotherapy helps to overcome movement avoidance due to fear of getting worse38,39. This shift in attitude, together with disability and pain improvements, may have contributed to the significantly increased levels of physical activity observed in the DG at the program-end. Indeed, strategies that promote healthy lifestyle habits, including physical activity, have long been recommended for CLBP9,10.
Work productivity is greatly affected by CLBP40, as observed in this cohort, with most participants reporting some level of absenteeism or presenteeism. It is known that physiotherapy can decrease absenteeism and presenteeism2,41. Herein, absenteeism was low at baseline and remained stable until program-end in both groups. However, absenteeism levels may be influenced by factors other than condition severity, such as financial support, work culture, secondary gain, coping skills, job satisfaction, and the ability to modify work42. Improvements in presenteeism were observed in both interventions without significant differences between groups.
Non-work-related activities can also be compromised in those suffering from CLBP43. In this study, both groups reported improvements in these activities, with no significant differences between them at the program-end, in line with current knowledge that physical activity can improve one’s ability to complete activities of daily living43.
Non-surgical treatments for CLBP have shown comparable or better outcomes in disability and return to work than surgical interventions2,12. Herein, the patient’s willingness to pursue surgery at baseline was low and decreased to near zero at the intervention-end, which is likely attributable to the reduction in pain and disability. This is auspicious considering that the patient’s surgery intention is one of the strongest predictors of future surgery44,45 and that many spinal surgeries are unnecessary2,5,12.
Opioid overuse and misuse represent a substantial problem in CLBP46. Opioid prescription rates have been reported to be twice as high as physiotherapy referral7, which contrasts with clinical recommendations9,10. A National survey conducted in the U.S. found that 21.5% of patients with chronic musculoskeletal pain take opioids7, while a cross-sectional study reported that 36.9% of American adults suffering specifically from CBLP use some type of analgesics8. In the present study, approximately half of the participants in both groups used analgesics, but only 10–13% reported opioid consumption at baseline. Both groups maintained a similar proportion of patients consuming analgesics until program-end despite the observed decreases in pain and disability, which is similar to a previous trial reporting no differences between in-clinic and web-based rehabilitation for LBP20. In future studies, higher granularity in medication assessment (i.e., type, dosage, and intake frequency) would help to better assess the effect of interventions in reducing pharmacological intake. Nevertheless, early physiotherapy has been reported to lower the risk of opioid misuse47,48.
In the context of insufficient resources and increasing demand for rehabilitation services1, alternative and scalable care delivery systems are needed. The growing body of evidence around digital care supports their use to overcome existing gaps15,16,17. These models will pave the way for a new integrated approach featuring digital, in-person, and hybrid care models. This will have profound cultural, clinical, and organizational implications for healthcare providers. Technology can be used to expand reach, promote timely access to care and provide critical data on patient progress and outcomes, but it requires appropriate candidate selection. Further research is required to determine the best candidates for each approach (digital, in-person or hybrid), considering clinical and demographic characteristics, and patient preference, as evidenced by changing attitudes towards digital care during the pandemic49. The widespread adoption of these new models requires the acquisition of new competencies by physiotherapists, requiring important changes in training curricula. Additionally, studies with long-term follow-ups and cost-effectiveness analyses are warranted.
A key strength of this study is the methodology — an RCT comparing a full DCP against high-intensity, in-person physiotherapy, with balanced treatment dosages between both groups. Of note, the treatment dosages used in both groups were much higher than typically utilized in U.S. outpatient clinics50. Other aspects of methodological rigor included randomization and allocation concealment, clinically relevant secondary outcomes, ITT analysis, reporting according to CONSORT guidelines, and trial registration. The novelty of the digital intervention is another strength of this study — a fully-remote multimodal DCP, asynchronously managed by a DPT, combining exercise with real-time biofeedback to education and CBT. The use of biofeedback guiding patients during sessions18,30 and the close monitoring and communication with the DPT may increase patient’s adherence17,28,29. This study is not without limitations, which include the lack of (i) blinding of patients, DPTs and investigators, due to the nature of the interventions; (ii) objective outcome measures (e.g., muscle strength, range of motion), including a metric for physical activity; and (iii) an externally blinded outcome adjudicator. Second, the study was conducted during and post-COVID-19 pandemic, which may have positively impacted compliance rates for both groups, more so in the digital intervention. Third, given the higher dropout rate in the CG, additional strategies to increase motivation and engagement could have been implemented, such as additional contacts between in-person sessions. Fourth, more granularity for the pharmacological variable (e.g., type, dosage, or frequency of analgesic intake) should be addressed in future studies. Finally, interventions performed under controlled environments may not reflect real-world conditions despite liberal selection criteria that yielded a population comparable to the reported in previous large-scale surveys and studies51.
In conclusion, this study supports the comparable effectiveness of a fully-remote digital intervention for CLBP patients compared to high-intensity, evidence-based, in-person physiotherapy. Similar improvements were noted in both groups, as well as high adherence and satisfaction, with lower dropout rates in the digital group. In the face of the high and growing burden of CLBP, these results support the consideration of digital interventions as viable and effective alternatives to in-person care, ensuring clinical quality and safety while reducing barriers to access.
This single-center, parallel-group, randomized controlled study was conducted in accordance with the Declaration of Helsinki. Following CONSORT guidelines, the trial was prospectively approved by the Emory Institutional Review Board (number STUDY00001546) and registered on ClinicalTrials.gov (NCT04808141) on March 22nd, 2021. Treatment occurred from June 15th, 2021, through October 26th, 2022.
Patients seeking care at the Emory Orthopedic and Spine Center (Atlanta, Ga) were screened for eligibility in-clinic by a physician who followed up participants until the study end. Informed written electronic consent was obtained from all participants (Castor eConsent, Castor Research Inc.). All study-related data was stored in an electronic data capture system (Castor EDC, Castor Research Inc.). Inclusion criteria were: (a) subjects between 18 and 80 years of age; (b) CLBP — intermittent or persistent LBP for at least 12 weeks, and/or ≥50% of the time in the past 6 months52; (c) ability to understand complex motor tasks; (d) ability to interact with a tablet. Exclusion criteria were: (a) known pregnancy; (b) spinal surgery <3 months ago; (c) symptoms and/or signs indicative of infection; (d) indication for spine surgery (i.e., tumor, cauda equina syndrome); (e) cancer diagnosis or undergoing cancer treatment; (f) known disorder incompatible with 20-minute light to moderate exercise; (g) concomitant non-spine related neurological disorder (e.g., stroke and multiple sclerosis); (h) dementia or psychiatric disorders precluding a patient from complying with a home-based exercise program; (i) illiteracy and/or serious visual or auditory impairment interfering with communication or compliance. Participants presenting with lumbar radicular pain were allowed to enter the study and were identified by: pain radiating below the buttock in a dermatomal distribution, altered sensation or paresthesia (e.g., hypoesthesia, numbness, and tingling), neurologic signs, or exam findings consistent with radicular pain53. To enroll in the study, participants were required to stop ongoing physiotherapy for CLBP.
After eligibility screening, participants who did not complete the assessment surveys or suffered serious adverse events54 were excluded. Participants were considered dropouts if they: (1) abandoned the study; or (2) did not engage in any exercise session for 28 consecutive days in the DG or missed 4 consecutive scheduled sessions in the CG.
Allocation and blinding
All participants providing consent were randomized into the DG or CG in a 1:1 ratio, using random permuted blocks of 4–8 participants, automatically generated by the Castor EDC platform. Group disclosure was only performed after randomization (concealed allocation), after which the principal investigator (D.C.) communicated the assignment to the study coordinator (L.G.). Blinding physiotherapists and patients to allocation was not possible, given the nature of the intervention.
The digital intervention group received an 8-week telerehabilitation intervention consisting of exercise complemented with education, and CBT delivered through a digital platform (Supplementary Table 5), which interfaced between the participant and the assigned DPT. DPTs involved in the study had, on average, 13 years of experience (range 6–23). An FDA-listed class II medical device comprising two inertial motion trackers, a mobile app on a dedicated tablet, and a cloud-based portal was made available to all participants. Hot spots were sent to participants without an internet connection. After an initial onboarding video call where the DPT assessed each participant, a tailored program was prescribed. Exercise sessions (three 20-minute sessions per week recommended32; a total of 24 sessions) were performed independently at the participants’ convenience through the tablet display. The system provided real-time biofeedback on performance through video-audio cues based on trackers placed with straps on the thoracic and lumbar regions (Supplementary Fig. 2).
Data on exercise session performance (range of motion, execution, and movement compensations) and engagement (number of executed or skipped repetitions, and time dedicated to exercise) was automatically recorded by the tablet app and stored in a cloud-based portal, enabling asynchronous monitoring by the assigned DPT, who performed adjustments to protocol.
The educational component consisting of articles focusing on anatomy and physiology, pain, exercise, and fear-avoidance behaviors9,10, was delivered through a smartphone app. A dedicated third-generation CBT program combining mindfulness, acceptance, commitment therapy, and empathy-focused therapy, adapted to a curriculum focused on chronic pain consisting of self-paced written and pre-recorded audio materials, was delivered through email9,10. Both education and CBT components were developed according to current guidelines9,10. Bi-directional communication was ensured by a built-in secure chat within a smartphone app or calls (with touchpoints scheduled every 4 weeks and on-demand), which was also intended to motivate and engage patients to the intervention. Members had access to technical and IT support through several communication channels. Hardware issues that could not be resolved remotely were solved by replacing the device.
The conventional intervention group received evidence-based in-person physiotherapy composed of exercise (following Emory’s standardized protocols: including strengthening, balance, stretching, and mobility exercises; similar to the digital group), education (regarding pain physiology, fear-avoidance behaviors and benefits/effects of exercise) on an ongoing and as-needed basis, manual therapy (e.g., joint mobilization and massage) and physical modalities (e.g., electrical stimulation) (Supplementary Table 6). DPTs involved in the study had, on average, 12 years of experience (range 7–20). The program was adjusted according to the participant’s condition and consisted of two 30-minute sessions per week for 8 weeks (total of 16 sessions)32. In specific circumstances, combining the two 30-minute sessions into one 60-minute session was allowed to circumvent unforeseen scheduling issues. Participants were also instructed to perform the exercises at home, but adherence to this was not assessed.
Outcomes were collected at baseline, 4, and 8 weeks (except physical activity levels at baseline and 8 weeks). Changes were calculated between baseline and 8 weeks.
The primary outcome was the Oswestry Disability Index (ODI) change between baseline and 8 weeks. ODI is validated for CLBP (with and without radiculopathy)55, including 10 items with a 6-point Likert scale (score: 0 (none)–100% (worse))56. The designation of a treatment responder was considered to be an MCID of 10-points or 30%57 at the primary 8-week endpoint. Secondary outcomes are presented in Table 5.
Safety and adverse events
CG sessions were performed under the direct supervision of a DPT. In the DG, exercise performance (motion trackers-based), as well as pain and fatigue data (faces-rating 5-item scale), were obtained following each session to support the PT asynchronous monitoring. Both participants and DPTs were instructed to contact the study investigators when adverse events occurred (registered on Castor EDC).
The sample size estimation was based on ODI (primary outcome). Considering the literature at the time of protocol submission, a standard deviation at a baseline of 17.83 was applied, supported by the study of Stankovic et al.58, who compared two different rehabilitation protocols for CLBP (N = 160). A difference between groups of 10-points was considered to be clinically meaningful based on the study by Ostelo et al.57. Considering a power of 80% and a two-sided 0.05 significance level, we calculated that 102 individuals would be necessary to detect a 10-point difference between the two groups. To guarantee that the study was adequately powered to detect equivalence, a posteriori analysis was conducted using the Two One-Sided Test (TOST) methodology59 (simulation-based power analysis). Assuming an 80% power to detect equivalence, a total of 104 participants (52 per arm) was estimated. A dropout rate of 15% was considered, which required to enroll 120 participants (60 per arm). To assess non-inferiority, a smaller sample size would be needed (total of 74; 37 per arm). To account for screening failures and dropouts, the participant enrollment period was extended beyond the original target.
Continuous data distribution was analyzed using the Kolmogorov–Smirnov test, followed by inspection of histograms and Q-Q plots. Baseline demographic and engagement metrics differences between groups were assessed using independent samples t-test or Mann–Whitney U test for quantitative variables and the Chi-square test for categorical variables. Baseline medication, opioid consumption, and the occurrence of adverse events at 8 weeks between groups were assessed through the Chi-square test.
The impact of each intervention on the primary and secondary outcomes was assessed using both the 8-week end scores and changes between baseline and 8 weeks. Outcomes were analyzed following both ITT, considering all randomized participants, and per-protocol analyses.
Since the assumption of normality was not met (Supplementary Fig. 1), logarithmic and Box–Cox transformations were performed. Results indicated that these transformations were not able to achieve normality (data not shown), and therefore, the repeated measures ANOVA analysis planned in the protocol was not conducted. As an alternative, a quantile mixed-effects model using a robust method on the medians was performed for the ITT analysis. This previously validated model accounts for repeated measurements and assumes non-normality60. Missing data were handled using multiple imputations by chained equations (MICE)61.
Considering group sizes in the per-protocol analysis (precluding quantile mix-effects model analysis), differences between interventions were evaluated using Mann–Whitney U tests with the Hodges–Lehmann estimator.
Considering the poorer prognosis associated with untreated radicular pain62, which affects care decision-making9,10,53, a subgroup analysis comparing clinical outcomes between those presenting with lumbar radicular pain and those without radiculopathy was performed using Mann–Whitney U tests.
Binary logistic regression analysis was performed to identify the odds of being a responder for ODI, considering a 10-points or 30% MCID57. This analysis was also conducted to assess the odds between groups for consuming analgesics, including opioids, at 8 weeks using the CG as a reference.
Physical activity was assessed using ordinal regression to evaluate whether the latent distribution of physical activity categories changed significantly from baseline to program-end within and between groups.
In all analyses, a two-sided hypothesis test with an alpha level of 0.05 was considered statistically significant. Robust linear mixed effects model and ordinal regression were coded using R (version 4.2.2, R Foundation for Statistical Computing) and all other analyses using SPSS (version 28.0, SPSS Inc, Chicago, Illinois, USA). Analysis of the data was performed by a blinded statistician.
The study will be available in clinicaltrials.gov upon publication for at least 5 years. Aggregated data of this study’s findings will be available upon reasonable request from the corresponding author.
Code for all statistical analyses may be available upon reasonable request from the corresponding author.
Cieza, A. et al. Global estimates of the need for rehabilitation based on the Global Burden of Disease study 2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet 396, 2006–2017 (2020).
Kamper, S. J. et al. Multidisciplinary biopsychosocial rehabilitation for chronic low back pain: cochrane systematic review and meta-analysis. Br. Med. J. 350, h444 (2015).
Lo, J., Chan, L. & Flynn, S. A systematic review of the incidence, prevalence, costs, and activity and work limitations of amputation, osteoarthritis, rheumatoid arthritis, back pain, multiple sclerosis, spinal cord injury, stroke, and traumatic brain injury in the United States: a 2019 update. Arch. Phys. Med. Rehabil. 102, 115–131 (2021).
Ganguli, I. et al. Cascade services and spending following low-value imaging for uncomplicated low back pain among commercially insured adults. J. Gen. Intern. Med. 38, 1102–1105 (2022).
Yanamadala, V. et al. Multidisciplinary evaluation leads to the decreased utilization of lumbar spine fusion: an observational cohort pilot study. Spine (Phila Pa 1976). 42, E1016–e1023 (2017).
Martin, B. I. et al. Trends in Lumbar Fusion Procedure Rates and Associated Hospital Costs for Degenerative Spinal Diseases in the United States, 2004 to 2015. Spine (Philos. Pa 1976). 44, 369–376 (2019).
Feldman, D. E., Carlesso, L. C. & Nahin, R. L. Management of patients with a musculoskeletal pain condition that is likely chronic: results from a national cross sectional survey. J. Pain 21, 869–880 (2020).
Shmagel, A. et al. Prescription medication use among community-based U.S. adults with chronic low back pain: a cross-sectional population based study. J. Pain 19, 1104–1112 (2018).
Qaseem, A. et al. Noninvasive treatments for acute, subacute, and chronic low back pain: a clinical practice guideline from the American College of Physicians. Ann. Intern. Med. 166, 514–530 (2017).
National Guideline Centre (UK). Low back pain and sciatica in over 16s: assessment and management. London: National Institute for Health and Care Excellence UK (NICE). https://www.nice.org.uk/guidance/ng59/resources/low-back-pain-and-sciatica-in-over-16s-assessment-and-management-pdf-1837521693637 (2016).
Hayden, J. A. et al. Exercise therapy for chronic low back pain. Cochrane Database Syst. Rev. 9, Cd009790 (2021).
Wang, L. et al. Surgical versus nonsurgical treatment of chronic low back pain: a meta-analysis based on current evidence. J. Back Musculoskelet. Rehabil. 29, 393–401 (2016).
Carvalho, E., Bettger, J. P. & Goode, A. P. Insurance coverage, costs, and barriers to care for outpatient musculoskeletal therapy and rehabilitation services. N. C. Med. J. 78, 312–314 (2017).
Collado-Mateo, D. et al. Key factors associated with adherence to physical exercise in patients with chronic diseases and older adults: an umbrella review. Int. J. Environ. Res. Public Health 18, 2023 (2021).
Snoswell, C. L., North, J. B. & Caffery, L. J. Economic advantages of telehealth and virtual health practitioners: return on investment analysis. JMIR Perioper. Med. 3, e15688 (2020).
Fatoye, F. et al. The clinical and cost-effectiveness of telerehabilitation for people with nonspecific chronic low back pain: randomized controlled trial. JMIR Mhealth Uhealth. 8, e15375 (2020).
Bennell, K. L. et al. Does a web-based exercise programming system improve home exercise adherence for people with musculoskeletal conditions?: A randomized controlled trial. Am. J. Phys. Med. Rehabil. 98, 850–858 (2019).
Chen, M. et al. Efficacy of mobile health in patients with low back pain: systematic review and meta-analysis of randomized controlled trials. JMIR Mhealth Uhealth. 9, e26095 (2021).
Corso, M. et al. Are Nonpharmacologic Interventions Delivered Through Synchronous Telehealth as Effective and Safe as In-Person Interventions for the Management of Patients With Nonacute Musculoskeletal Conditions? A Systematic Rapid Review. Arch. Phys. Med Rehabil. 103, 145–154.e111 (2022).
Raiszadeh, K. et al. In-clinic versus web-based multidisciplinary exercise-based rehabilitation for treatment of low back pain: prospective clinical trial in an integrated practice unit model. J. Med. Internet Res. 23, e22548 (2021).
Toelle, T. R. et al. App-based multidisciplinary back pain treatment versus combined physiotherapy plus online education: a randomized controlled trial. npj Digit. Med. 2, 34 (2019).
Correia, F. D. et al. Medium-term outcomes of digital versus conventional home-based rehabilitation after total knee arthroplasty: prospective, parallel-group feasibility study. JMIR Rehabil. Assist. Technol. 6, e13111 (2019).
Dias Correia, F. et al. Digital versus conventional rehabilitation after total hip arthroplasty: a single-center, parallel-group pilot study. JMIR Rehabil. Assist. Technol. 6, e14523 (2019).
Costa, F. et al. Digital rehabilitation for acute low back pain: a prospective longitudinal cohort study. J. Pain. Res. 15, 1873–1887 (2022).
Areias, A. C. et al. Impact on productivity impairment of a digital care program for chronic low back pain: a prospective longitudinal cohort study. Musculoskelet. Sci. Pract. 63, 102709 (2022).
Joseph, R. et al. A systematic review finds variable use of the intention-to-treat principle in musculoskeletal randomized controlled trials with missing data. J. Clin. Epidemiol. 68, 15–24 (2015).
Mbada, C. E. et al. Comparative Efficacy of Clinic-Based and Telerehabilitation Application of Mckenzie Therapy in Chronic Low-Back Pain. Int J. Telerehabil. 11, 41–58 (2019).
Argent, R., Daly, A. & Caulfield, B. Patient involvement with home-based exercise programs: can connected health interventions influence adherence? JMIR Mhealth Uhealth. 6, e47 (2018).
Miciak, M. et al. The necessary conditions of engagement for the therapeutic relationship in physiotherapy: an interpretive description study. Arch. Physiother. 8, 3 (2018).
Cranen, K. et al. Toward patient-centered telerehabilitation design: understanding chronic pain patients' preferences for web-based exercise telerehabilitation using a discrete choice experiment. J. Med. Internet Res. 19, e26 (2017).
Unsgaard-Tøndel, M. & Søderstrøm, S. Therapeutic alliance: patients' expectations before and experiences after physical therapy for low back pain-a qualitative study with 6-month follow-up. Phys. Ther. 101, pzab187 (2021).
Mueller, J. & Niederer, D. Dose-response-relationship of stabilisation exercises in patients with chronic non-specific low back pain: a systematic review with meta-regression. Sci. Rep. 10, 16921 (2020).
Daré, L. O. et al. Co-morbidities of mental disorders and chronic physical diseases in developing and emerging countries: a meta-analysis. BMC Public Health 19, 304 (2019).
Crofford, L. J. Psychological aspects of chronic musculoskeletal pain. Best. Pract. Res. Clin. Rheumatol. 29, 147–155 (2015).
Stewart Williams, J. et al. Risk factors and disability associated with low back pain in older adults in low- and middle-income countries. Results from the WHO Study on Global AGEing and Adult Health (SAGE). PLoS ONE 10, e0127880 (2015).
Bossen, D. et al. Effectiveness of a web-based physical activity intervention in patients with knee and/or hip osteoarthritis: randomized controlled trial. J. Med. Internet Res. 15, e257 (2013).
Zheng, F. et al. The Effect of M-Health-Based Core Stability Exercise Combined with Self-Compassion Training for Patients with Nonspecific Chronic Low Back Pain: A Randomized Controlled Pilot Study. Pain Ther. 11, 511–528 (2022).
Wertli, M. M. et al. The role of fear avoidance beliefs as a prognostic factor for outcome in patients with nonspecific low back pain: a systematic review. Spine J. 14, 816–836.e814 (2014).
Martinez-Calderon, J. et al. Conservative interventions reduce fear in individuals with chronic low back pain: a systematic review. Arch. Phys. Med. Rehabil. 101, 329–358 (2020).
Fullen, B. et al. Management of chronic low back pain and the impact on patients’ personal and professional lives: results from an international patient survey. Pain. Pract. 22, 463–477 (2022).
Ibrahim, M. E. et al. Recovering the capability to work among patients with chronic low Back pain after a four-week, multidisciplinary biopsychosocial rehabilitation program: 18-month follow-up study. BMC Musculoskelet. Disord. 20, 439 (2019).
Wynne-Jones, G. et al. Absence from work and return to work in people with back pain: a systematic review and meta-analysis. Occup. Environ. Med. 71, 448–456 (2014).
Zhang, S. K. et al. Effects of low back pain exercises on pain symptoms and activities of daily living: a systematic review and meta-analysis. Percept. Mot. Skills 129, 63–89 (2022).
Modi, C. S. et al. Factors that influence the choice to undergo surgery for shoulder and elbow conditions. Clin. Orthop. Relat. Res. 472, 883–891 (2014).
Hawker, G. A. et al. A prospective population-based study of the predictors of undergoing total joint arthroplasty. Arthritis Rheum. 54, 3212–3220 (2006).
Tucker, H. R. et al. Harms and benefits of opioids for management of non-surgical acute and chronic low back pain: a systematic review. Br. J. Sports Med. 54, 664 (2020).
Sun, E. et al. Association of early physical therapy with long-term opioid use among opioid-naive patients with musculoskeletal pain. JAMA Netw. Open. 1, e185909 (2018).
Martin, S. et al. Does early physical therapy intervention reduce opioid burden and improve functionality in the management of chronic lower back pain? Ment. Health Clin. 10, 215–221 (2020).
King, D. et al. Transformation from a traditional model to a virtual model of care in orthopaedic surgery: COVID-19 experience and beyond. Bone Jt Open 1, 272–280 (2020).
Bhavsar, N. A. et al. Prevalence and predictors of no-shows to physical therapy for musculoskeletal conditions. PLoS ONE 16, e0251336 (2021).
Shmagel, A., Foley, R. & Ibrahim, H. Epidemiology of chronic low back pain in US adults: data from the 2009-2010 National Health and Nutrition Examination Survey. Arthritis Care Res. 68, 1688–1694 (2016).
Deyo, R. A. et al. Report of the NIH Task Force on Research standards for chronic low back pain. J. Pain 15, 569–585 (2014).
Khorami, A. K. et al. Recommendations for diagnosis and treatment of lumbosacral radicular pain: a systematic review of clinical practice guidelines. J. Clin. Med. 10, 2482 (2021).
U.S. Food and Drug Administration. What is a Serious Adverse Event? https://www.fda.gov/safety/reporting-serious-problems-fda/what-serious-adverse-event (2016).
Saltychev, M. et al. Psychometric properties of the Oswestry Disability Index. Int. J. Rehabil. Res. 40, 202–208 (2017).
Fairbank, J. C. & Pynsent, P. B. The Oswestry Disability Index. Spine (Phila Pa 1976). 25, 2940–2953 (2000).
Ostelo, R. W. et al. Interpreting change scores for pain and functional status in low back pain: towards international consensus regarding minimal important change. Spine (Phila Pa 1976). 33, 90–94 (2008).
Stankovic, A. L. M. et al. Lumbar stabilization exercises in addition to strengthening and stretching exercises reduce pain and increase function in patients with chronic low back pain: randomized clinical open-label study. Turk. J. Phys. Med. Rehabil. 58, 177–183 (2012).
Lakens, D. Equivalence tests: a practical primer for t tests, correlations, and meta-analyses. Soc. Psychol. Personal. Sci. 8, 355–362 (2017).
Liu Y & Bottai M. Mixed-effects models for conditional quantiles with longitudinal data. Int. J. Biostat. 5, Article 28 (2009).
White, I. R., Royston, P. & Wood, A. M. Multiple imputation using chained equations: issues and guidance for practice. Stat. Med. 30, 377–399 (2011).
Konstantinou, K. et al. The impact of low back-related leg pain on outcomes as compared with low back pain alone: a systematic review of the literature. Clin. J. Pain. 29, 644–654 (2013).
Childs, J. D., Piva, S. R. & Fritz, J. M. Responsiveness of the numeric pain rating scale in patients with low back pain. Spine (Phila Pa 1976). 30, 1331–1334 (2005).
George, S. Z., Valencia, C. & Beneciuk, J. M. A psychometric investigation of fear-avoidance model measures in patients with chronic low back pain. J Orthop Sports Phys Ther. 40, 197–205 (2010).
Spitzer, R. L., Kroenke, K., Williams, J. B. & Löwe, B. A brief measure for assessing generalized anxiety disorder: the GAD-7. Arch Intern Med. 166, 1092–1097 (2006).
Bijker, L., Sleijser-Koehorst, M. L. S., Coppieters, M. W., Cuijpers, P. & Scholten-Peeters, G. G. M. Preferred Self-Administered Questionnaires to Assess Depression, Anxiety and Somatization in People With Musculoskeletal Pain - A Modified Delphi Study. J Pain. 21, 409–417 (2020).
Kroenke, K., Spitzer, R. L. & Williams, J. B. The PHQ-9: validity of a brief depression severity measure. J Gen Intern Med. 16, 606–613 (2001).
Lee, P. H., Macfarlane, D. J., Lam, T. H. & Stewart, S. M. Validity of the International Physical Activity Questionnaire Short Form (IPAQ-SF): a systematic review. Int J Behav Nutr Phys Act. 8, 115 (2011).
IPAQ group. Guidelines for data processing and analysis of the International Physical Activity Questionnaire (IPAQ). https://sites.google.com/site/theipaq/scoring-protocol (2005).
Ospina, M. B., Dennett, L., Waye, A., Jacobs, P. & Thompson, A. H. A systematic review of measurement properties of instruments assessing presenteeism. Am J Manag Care. 21, e171–185 (2015).
The authors acknowledge the team of physiotherapists responsible for the management of participants: Stacie Segebart, Abby Shottler, Ricky Schrantz, Pamela Gjerde, and Erika Ortega. We also would like to acknowledge the clinical research coordination team: Isabelle M. Vernon, Lauren Glenney, Diya Sandhu, Mimi Nguyen, and David Pasion. No funding was obtained for this work.
The authors declare no competing interests but the following competing financial interests: F. Costa, D.J., A.C.A., M.M., F.D.C., and V.Y. are employees of Sword Health, the sponsor of this study. F.D.C., V.Y., and V.B. also hold equity in Sword Health. R.M. is an independent scientific consultant responsible for statistical analysis. J.K.S. and S.P.C. are independent scientific and clinical consultants who received an adviser honorarium from Sword Health.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
About this article
Cite this article
Cui, D., Janela, D., Costa, F. et al. Randomized-controlled trial assessing a digital care program versus conventional physiotherapy for chronic low back pain. npj Digit. Med. 6, 121 (2023). https://doi.org/10.1038/s41746-023-00870-3