Patient-friendly summary
If you read nothing else
Bottom line
A carefully designed, pre-registered trial protocol for an AI-tailored self-management app for low back pain; it sets up a promising test but reports no results, so it cannot yet tell us whether the app is effective.
Moderate evidencePublished
Evidence hierarchy
Study participants
Danish and Norwegian adults (18+) with nonspecific low back pain who sought primary care within the previous 8 weeks (RMDQ ≥6)
Study Summary
This is the published protocol for the selfBACK trial, a single-blinded, two-arm randomized controlled trial designed in Denmark and Norway to test a smartphone app that supports self-management of nonspecific low back pain. The app delivers weekly tailored self-management plans — physical activity (step) goals, strength and flexibility exercises, and educational content — individualized using case-based reasoning, a branch of artificial intelligence. The trial planned to recruit 350 adults who had sought primary-care help for low back pain within the preceding 8 weeks, randomizing them to the app plus usual care versus usual care alone, with pain-related disability (Roland-Morris Disability Questionnaire) at 3 months as the primary outcome. As a protocol, it describes the study design and rationale only; results of the trial were expected to be published in fall 2020 and are not reported here.
Key Findings
| Finding | Detail | Impact |
|---|---|---|
| This is a trial protocol — no outcome results are reported yet | The paper describes the planned design, methods, and rationale of the selfBACK RCT. Recruitment opened in February 2019; data collection was expected to complete by fall 2020 and primary-outcome results were expected to be published thereafter. No effectiveness data are presented in this document. | High |
| The app tailors weekly self-management plans using case-based reasoning (artificial intelligence) | Plans combine recommendations on daily step count, a program of strength and flexibility exercises, and educational material. Tailoring draws on the baseline questionnaire, a weekly app-based tailoring session, the participant's reported exercise completion, and step counts from a wristband (Mi Band 3). The case-based reasoning system reuses solutions from previous similar participant cases to individualize advice. | High |
| The primary outcome is pain-related disability at 3 months, with a 2-point RMDQ difference targeted | The primary end point is the difference in Roland-Morris Disability Questionnaire (RMDQ) score between the app-plus-usual-care group and the usual-care-only group at 3 months. The trial aimed to detect a 2-point between-group difference; the authors note a 5-point difference has been reported as clinically important by some, while others suggest a 1- to 2-point difference may be important when baseline disability is low. | Medium |
| Content was theory-driven and evidence-based, drawing on an intervention-mapping process and prior testing | The app content was developed via intervention mapping and refined through two feasibility studies and one pilot study before the RCT version was finalized. The exercise bank held 56 strength and flexibility exercises plus 14 pain-relief exercises; educational material was organized under 14 main categories. The authors note prior commercial LBP apps were largely of poor quality and untested, framing this as a key knowledge gap. | Medium |
| Self-management is a recommended first-line treatment but its effects in LBP have been modest | The authors cite systematic reviews reporting self-management effects on pain as moderate and on pain-related disability as small to moderate, attributed partly to wide variation in program content and poor adherence. Tailoring and digital delivery are proposed as ways to improve engagement and effectiveness. | Medium |
The paper describes the planned design, methods, and rationale of the selfBACK RCT. Recruitment opened in February 2019; data collection was expected to complete by fall 2020 and primary-outcome results were expected to be published thereafter. No effectiveness data are presented in this document.
Plans combine recommendations on daily step count, a program of strength and flexibility exercises, and educational material. Tailoring draws on the baseline questionnaire, a weekly app-based tailoring session, the participant's reported exercise completion, and step counts from a wristband (Mi Band 3). The case-based reasoning system reuses solutions from previous similar participant cases to individualize advice.
The primary end point is the difference in Roland-Morris Disability Questionnaire (RMDQ) score between the app-plus-usual-care group and the usual-care-only group at 3 months. The trial aimed to detect a 2-point between-group difference; the authors note a 5-point difference has been reported as clinically important by some, while others suggest a 1- to 2-point difference may be important when baseline disability is low.
The app content was developed via intervention mapping and refined through two feasibility studies and one pilot study before the RCT version was finalized. The exercise bank held 56 strength and flexibility exercises plus 14 pain-relief exercises; educational material was organized under 14 main categories. The authors note prior commercial LBP apps were largely of poor quality and untested, framing this as a key knowledge gap.
The authors cite systematic reviews reporting self-management effects on pain as moderate and on pain-related disability as small to moderate, attributed partly to wide variation in program content and poor adherence. Tailoring and digital delivery are proposed as ways to improve engagement and effectiveness.
Strengths
- Rigorous, transparent design: pre-registered (ClinicalTrials.gov NCT03798288), intention-to-treat analysis, reporting following SPIRIT and CONSORT-EHEALTH guidelines, and a blinded two-interpretation procedure to reduce biased interpretation of results.
- App content was theory-driven and evidence-based, developed through an intervention-mapping process and refined in two feasibility studies and a pilot study before the trial.
- Pragmatic, real-world design recruiting a general care-seeking low back pain population across two countries, giving high external validity.
- Includes a parallel mixed-methods process evaluation (guided by Normalization Process Theory and the RE-AIM framework) to understand uptake, use, and implementation.
Limitations
- This is a protocol only — it reports no effectiveness, safety, or outcome data; the actual trial results are not contained in this paper.
- The trial is single-blinded; participants are aware of their group allocation, which can influence self-reported outcomes such as disability.
- Usual care (the comparator) varies within and across the two countries and is not standardized, making the precise comparison condition heterogeneous.
- The multi-component design (exercise, physical activity, education) means that even if the app is shown to be effective, the trial cannot determine which component drives any benefit.
- Findings would apply to a smartphone-owning, internet-using, care-seeking Scandinavian population and may not generalize to other settings or to people without digital access.
Key Takeaways for Patients
What This Means for You
- 01Self-management — staying active, doing back exercises, and learning about your condition — is recommended as a first-line approach for nonspecific low back pain.
- 02This study tested an app that builds a personalized weekly plan combining a daily step goal, strength and flexibility exercises, and educational tips, individualized by software that learns from similar past cases.
- 03The app was designed to add to, not replace, the care from your own clinician; participants were told to keep following their health professional's advice and seek care if symptoms worsened.
- 04Because this is a study protocol, it does not yet tell us whether the app actually worked — those results were planned for a later publication.
- 05Many low back pain apps on app stores have been found to be poor quality and untested, so it is worth being cautious about commercial apps that have not been independently evaluated.
Read the Full Paper
Access the complete peer-reviewed study from JMIR Research Protocols
View Full Study