Inclusion and exclusion criteria are outlined in Textbox 1. The assessment of whether the criteria are considered to limit participation is performed either by the referring health care professional (HCP) or based on participant’s self-report. The selfBACK intervention is tested on a general LBP population rather than a specific subgroup to reflect that the intervention targets care-seeking patients not limited to specific characteristics such as symptom duration.

Recruitment is performed in Trondheim, Norway, and Odense, Denmark. The recruitment flow is described in Figure 1. A total of 350 participants are to be recruited to the RCT. Of these, 75% (262/350) will be recruited in Denmark and 25% (88/350) in Norway. Recruitment is undertaken by physiotherapists, chiropractors, and general practitioners. In Denmark, participants are additionally recruited from the Spine Centre of Southern Denmark, an outpatient hospital that provides care for people with back pain referred from primary care, either family physicians or chiropractors. The Spine Centre provides diagnostic assessment and prescribes treatment plans. For all recruitment sites, people seeking care due to nonspecific LBP may be referred to the trial by the consulting HCP based on a short description of eligibility for the trial. Final eligibility is assessed by the research team during a screening phone call. The recruitment to the selfBACK trial will not affect any planned routine diagnostic assessment or treatment (usual care).

Interested patients are screened via telephone by a member of the research team. If eligible and willing to participate, participants give their verbal consent to participate and are invited to complete the baseline questionnaire. Thereafter participants give their written consent to participate and are randomized to one of two groups.

Participants are randomized to either (1) selfBACK DSS in addition to usual care or (2) usual care only. Randomization is performed as a block randomization with permuted blocks of random size and stratified by country and care provider (ie, general practitioner, physiotherapist, chiropractor, or Spine Centre). The allocation ratio between groups is 1:1. Randomization is performed in a Web-based trial management system (Web Case Report Form [WebCRF]) developed and administered by the Unit of Applied Clinical Research, Faculty of Medicine and Health Sciences, Norwegian University of Science and Technology (NTNU), Trondheim, Norway. The WebCRF system holds a minimal dataset on all screened participants (variables include trial indentification number, participant initials, country, type of HCP recruiting participant, age, and gender). The study is single-blinded; participants are not blinded to group allocation. Analysis and interpretation of the study results will be performed by researchers blinded to group allocation.

The trial and intervention are described following the Standard Protocol Items: Recommendations for Interventional Trials [24] and Consolidated Standards of Reporting Trials of Electronic and Mobile Health Applications and Online Telehealth (CONSORT-EHEALTH) [25] guidelines.

The primary outcome is pain-related disability at 3-month follow-up assessed using the RMDQ [29]. The questionnaire includes 24 items asking participants to indicate if they experience functional impairments by answering yes or no to a series of descriptions of functional abilities. Higher scores indicate higher level of disability [30]. For the selfBACK trial, we aim to identify a 2-point difference in RMDQ between the intervention and control group at 3-month follow-up. The rationale for selecting this cutoff was based on several considerations. First, self-management through selfBACK is included as an add-on to usual care in this trial. Although the magnitude of effect for this novel intervention is difficult to predict, a small beneficial effect above that of usual care could be important for this group of patients. Second, the suggested minimal clinically important difference in RMDQ may vary according to the disability level in the population under study [31]. Even though a 5-point difference has been reported as clinically important [32], others have suggested a 1- to 2-point difference to be clinically important if the disability level is low [33].

Descriptive variables include age, gender, height, weight, and report of any comorbidities (comorbidities were registered using an existing questionnaire (HUNT3) from the Norwegian HUNT study [34]). Demographic variables including family relations, ethnicity, educational status, employment, and work characteristics if employed are collected at baseline (Table 2).

A range of secondary and exploratory outcomes is included in the trial, and participants randomized to use the selfBACK app in addition to usual care are asked a set of tailoring questions weekly to individualize their self-management plan (Table 2). App use data such as number of visits, duration spent using the app, achievement scores, and number of days with visits are registered (Matomo software).

Outcome measures are collected at baseline, 6 weeks and 3, 6, and 9 months. Data collection is Web-based, and all data are entered directly into the selfBACK database by the participants. To maximize response rate, reminder emails are sent after 3 days and again after 6 days if no response to the first email. If still no answer, a researcher will contact the participant and ask if they are willing to answer the RMDQ questionnaire over the phone at follow-ups.

All outcome and other data are stored on secure servers at NTNU, the servers are firewall protected, and back-up is performed daily. Data storage is consistent with national (Denmark and Norway) and European regulations on data protection. Also, all data transferring processes are protected using https and Secure Sockets Layer as well as sending the data in encrypted format.

The sample size calculations have been performed in two ways. First, we performed a calculation assuming only one follow-up measure and a standard deviation of the RMDQ score of 6 points. The expected standard deviation was informed by previous high-quality studies in Denmark and United Kingdom investigating similar LBP populations [47-50]. Based on this calculation, we estimated that a sample size of 382 (191 in each group) was necessary to detect a 2-point difference with 90% power and a 2-sided alpha level of .05.

Second, we performed a simulation using 1000 repetitions of a mixed-model regression for repeated measures assuming (1) 3 data points per participant (ie, baseline, 6 weeks, and 3 months), (2) a 2-point difference between groups on RMDQ at 3 months, (3) a standard deviation of 6 points, and (4) a correlation between repeated measures of 0.4. The latter was based on information from previous trials with repeated measures for the RMDQ in similar LBP populations [51,52]. Based on these assumptions and an alpha level of .05, sample size calculations show that 250 participants (ie, 125 participants in each group) result in a power of 92% (95% CI 90%-93%) to detect a 2-point difference in RMDQ between the intervention group and control group at 3 months. Furthermore, simulations assuming a 2-point difference between groups observed at both follow-up time points (6 weeks and 3 months) indicated that a sample size of 180 (90 in each group) will result in a power of 94% (95% CI 92%-95%). These sample size calculations indicate that a sample size of approximately 250 persons (125 in each group) is adequate when using the repeated measure design. A recent systematic review showed that attrition rates ranged between 4% to 94% for digital self-management interventions lasting between 2 weeks and 12 months in LBP populations [22]. To allow for a 30% dropout rate at 3-month follow-up, we aim to include a total of 350 participants in the trial (175 participants in each arm).

The primary analysis will estimate mean group difference with 95% confidence interval of the RMDQ score over the first 3 months. Analyses will be conducted according to the intention-to-treat principle using a linear mixed model for repeated measures. This model includes all available data for all participants at each time point (ie, baseline, 6 weeks, and 3 months). In the regression model, individual participants will be specified as a random effect, accounting for the within-subject covariance structure. The effect of group and time will be specified as fixed effects using a joint variable of intervention and time. The analysis will investigate the effect of the intervention as constant over time, as well as an interaction between time and group allocation. Here, baseline levels are pooled over the two study groups assuming that any baseline differences are due to chance [53]. All effects will be estimated both crude and adjusted for the two variables used for stratification in the randomization (ie, country and care provider) [54]. Any missing values are inherently accounted for in the mixed-model approach [55].

To increase transparency, a statistical analysis plan will be agreed upon and made publicly available before ending the collection of the primary outcome. To reduce the risk of biased interpretation of results, the following procedure will be undertaken: two interpretations will be drafted based on a review of the primary outcome data with groups arbitrarily labeled A and B [56]. One interpretation assumes that A is the intervention group and B the control group, the other interpretation assumes the reverse. After agreeing on both interpretations, the randomization code is broken and the correct interpretation chosen.

A process evaluation exploring how participants use the intervention in daily life will be conducted as an integrated part of the RCT. For this we will use a mixed-methods process evaluation: gathering quantitative measures by questionnaires for participants including the Virtual Care Climate Questionnaire [57] and 3 rating questions (overall rating of the app, ease of use, recommendable to others), measures of data analytics on app use, and semistructured qualitative interviews. Normalization process theory [58], an implementation theory used extensively to identify barriers and facilitators to uptake and use of new technologies [59], will provide the conceptual underpinning to the process evaluation. The process evaluation will be guided by the RE-AIM framework and investigate all 5 elements of the framework: reach, effectiveness, adoption, implementation, and maintenance [60]. The full details on design and methods for the process evaluation will be published separately.

The trial was approved by the national ethical committees in Denmark (S-20182000-24) and Norway (2017/923-6). Correspondingly, national review boards or data protection agencies have approved the trial. In Denmark, approval was granted from the Danish Data Protection Agency through application to the University of Southern Denmark’s legal office (201-57-0008) and in Norway from the National Data Protection Authority or the Centre for Research Data through the ethics approval. The trial is registered with ClinicalTrials.gov [NCT03798288].

The trial results will be reported in accordance with the CONSORT 2010 reporting guideline and the 2013 CONSORT-EHEALTH checklist amendment for reporting Web-based and mobile-based RCTs [25,61].

No serious adverse events are expected for this trial. Should a participant contact the research team concerning any worsening of symptoms, the participant will be advised to seek care from their HCP as they normally would. All inquiries regarding potential adverse events will be recorded and discussed in an internal audit and reported with the study results. In addition, the selfBACK DSS is designed to react to increased pain or deterioration in symptoms, and it will adjust the self-management plans based on this information. In addition, participants are informed in the written information and during the screening call and inclusion process that this intervention is an add-on to usual care and should not replace contact with their HCP and that they should always follow the advice of the consulting HCP. Also, the app contains a Caution section describing worsening in symptoms that should be acted upon and advising participants to seek care from their consulting HCP if they experience any such symptoms.