This randomized, prospective, controlled study was a single-blind trial, conducted at the Department of Rheumatology and Immunology, Chinese People’s Liberation Army (PLA) General Hospital. After baseline assessments by an assessor who was blinded to the group assignments, eligible participants were randomly assigned to the trial group or the control group after providing informed consent. Assessors were blinded to group allocation, but participants were aware of their assigned intervention (digital platform or manual-based exercise) due to the inherent nature of the interventions. This pragmatic design reflects real-world conditions where patients would typically know their treatment modality. This protocol is in line with CONSORT (Consolidated Standards of Reporting Trials) guidelines (checklist in Multimedia Appendix 1).

Patients diagnosed with AS according to the 1984 modified New York criteria [7] at the Chinese PLA General Hospital, either as a first-time diagnosis or previously, were enrolled in the study and assigned to 2 groups. In addition to collecting their demographic characteristics (age, gender, weight, height, and BMI), the patients were also questioned about occupation, main symptoms, time of diagnosis, and drug usage (nonsteroidal anti-inflammatory drug, disease-modifying antirheumatic drugs, including biologics). Other inclusion criteria were aged ≥18 years, stable drug treatment in the preceding month, and ASDAS-CRP between 1.3 and 3.5 [8]. Exclusion criteria were cardiovascular disease or clinical status at high risk, screened with the American Heart Association/ACSM Health/Fitness Facility Preparticipation Screening Questionnaire [9], cervical vertebra bridges, surgery within the preceding 6 months, regular exercise in the preceding 3 months (eg, yoga, Tai Chi, Baduanjin 3 or more times per week, 20 minutes per time), and factors leading to the inability to receive regular exercise rehabilitation (such as language impairment, difficulty in understanding, and limited movements).

All subjects in this study were patients diagnosed with AS in the outpatient Department of Rheumatology and Immunology, in the Chinese PLA General Hospital. First, the potentially eligible subjects were screened by a study physician, and the exclusion criteria of all screened patients were recorded. Then, the informed consent form was signed for eligible subjects, and the baseline data were collected in detail. The trial implemented evaluation physician blinding at both baseline and follow-up phases, with group allocation data concealed via an opaque envelope. Postassessment allocation to a physical therapist for exercise prescription occurred through a separate workflow managed by the trial coordinator. Finally, a personalized exercise rehabilitation scheme was given to all the subjects by the physical therapist according to the patient’s function evaluation (Multimedia Appendix 2).

A 24-week tailored functional exercise program, consisting of stretching, mobilization, and strengthening exercises, was designed based on the patients’ basic functional abilities following randomization. All patients had a 1-hour training session at entry, providing information about AS and the purpose of physical exercises.

In addition to a face-to-face session, participants in the experimental group received a digital exercise program given by the physiotherapist. Participants in the intervention group were taught the usage of the digital exercise project that will be delivered via WeChat applets to individual exercise interfaces. In the first step, patients needed to wear a sensor on their chest, close to their armpits, and then connect to their mobile phones via Bluetooth. After a successful connection, they carried out the personalized workout program listed in Multimedia Appendix 2. During exercise, participants in the experimental group received real-time feedback on movement standardization and joint angles via the WeChat-based exercise applets. Postexercise feedback was systematically collected using the Borg Rating of Perceived Exertion scale (Figure 1), enabling adjustments to exercise regimens based on participants’ subjective experiences of effort.

The physiotherapist’s management platform enabled remote monitoring of participants’ home exercise compliance through real-time data transmission from wearable sensors. Participants were instructed to complete exercises 3-4 times per week, with session frequency and duration tailored to their baseline functional assessments. To reinforce adherence, automated reminders were triggered via the platform if a participant missed exercise sessions for 7 consecutive days, prompting re-engagement through personalized notifications (eg, motivational messages and corrective feedback). Physiotherapists reviewed compliance metrics weekly to adjust exercise prescriptions or provide additional guidance as needed.

The participants in the control group received a home-based exercise program, and an exercise manual was given to each participant. The participants in the control group were also recommended to perform the exercise 3-4 times per week. Participants in the control group were provided with a paper logbook to record exercise frequency, duration, and perceived exertion. Research assistants conducted biweekly phone follow-ups to collect logbook data, reinforce adherence, and address participant questions. Adherence was calculated as the self-reported ratio of completed sessions to prescribed sessions (3-4 sessions/wk), mirroring the intervention group’s digitally recorded adherence metric.

To minimize deviation risks, we established WeChat patient management groups for each participant (both control and intervention groups) comprising: the enrolled patient, a rheumatologist, a physical therapist, and a digital health assistant. The physical therapist provided targeted assistance in the intervention group under two predefined conditions: (1) when abnormal exercise difficulty was detected through our background monitoring system, and (2) if the intervention group participants failed to complete exercises consecutively for 1 week.

For participants in the control group, physical therapists solely addressed inquiries related to medical conditions and exercise clarification, without providing proactive rehabilitation guidance or progress monitoring. All communication occurred digitally via the WeChat platform except during scheduled in-person outpatient follow-up visit. No additional face-to-face meetings were conducted beyond these standard clinical appointments.

The digital exercise platform, Healbone, was developed through a participatory design process involving patients with AS and their relatives, rheumatology specialists, physical therapists, and a technical team. Healbone serves as a shared communication platform accessible to patients, their families, and associated health care professionals. Patients must provide written consent regarding data sharing. All information is uniformly available to all participants. Health care personnel and researchers have access to additional tools for managing care, obtaining overviews of patient conditions, and simultaneously monitoring multiple patients. The platform also functions as a coordination and communication tool among patients, relatives, and health care professionals.

All measurements are securely stored in a database. The web-based Healbone platform includes multiple features designed and developed through user-driven innovation. Facilities and input data specific to individual patients, such as symptom scores, are accessible and available to authorized personnel. The main features include:

In addition, Healbone can integrate with electronic health record systems and other medical devices, enabling comprehensive data management and analysis to provide holistic health management support for patients. Through these features, the platform not only improves patients’ physical abilities and quality of life but also equips clinicians with effective tools for enhanced monitoring and treatment plan adjustments (Figure 2).

All patients were recruited by a physician and randomly assigned to either the intervention or control group in a 1:1 ratio. To ensure allocation concealment, a computer-generated randomized table of numbers was prepared by an independent individual before the study’s commencement. The numbered cards indicating the random assignment and group allocation details were sealed in opaque envelopes by the same independent person. The physical therapist then opened these envelopes and administered the exercise program based on the assigned group.

For a schedule of data collection and measures list, refer to Table 1.

Statistical tests will be performed with SPSS Statistics (version 29; IBM Corp). Significance is defined as P<.05. Quantitative data are reported as means and SD. Statistical significance for quantitative variables is assessed using Student t test or 1-way ANOVA. Categorical data are presented as percentages, and significant differences are determined using chi-square tests.

The sample size was calculated using ASDAS-CRP as the primary outcome. Based on a previous randomized controlled trial comparing supervised versus home-based exercise in patients with AS, the study of the effect size was 0.714 [6]. Assuming a 2-tailed independent t test with α=.05, power 1−β=.8, and a 1:1 allocation ratio, G*Power 3.1 estimated a required sample size of 32 participants per group (total N=64). To account for a conservative 20% attrition rate-consistent with dropout rates reported in comparable AS exercise trial, we increased the total sample size to N=80 (40 per group).

Missing data due to participant dropout will be addressed using the Last Observation Carried Forward (LOCF) method for the primary analysis. Under this approach, missing values at the 24-week follow-up will be replaced with the last available measurement from earlier timepoints. This method assumes that unobserved outcomes remain stable after dropout, a rationale consistent with prior studies in chronic conditions like AS, where short-term interventions may yield sustained effects. All randomized and consented patients, constituting the full analysis set, are going to be included according to intention-to-treat principles for the efficacy analyses.

This study was approved by the Ethics Committee of the Chinese PLA General Hospital (number S2022-738-02) in compliance with the Declaration of Helsinki. All participants provided written consent after being informed of the study’s goals, procedures, risks, and benefits. They retained the right to withdraw at any time without affecting their care. Participant data were anonymized and securely stored on the Healbone platform using encryption, two-factor authentication, and restricted access. Personal details (eg, names) were stored separately. No financial payment was provided. Participants received standard clinical care and exercise programs for AS. No identifiable images or personal information appeared in this paper. Any future inclusion of such content will require explicit written consent.

This study anticipates that the digital functional exercise program (Healbone), which combines wearable sensors and real-time feedback, will lead to greater reductions in disease activity (ASDAS-CRP) and improvements in physical function (BASFI and BASMI) compared to standard home-based exercises. These benefits are expected to arise from higher adherence rates in the intervention group, driven by the platform’s reminders and corrective guidance. Existing research consistently shows that supervised exercise programs outperform home-based regimens in managing AS but face practical challenges like cost and accessibility [14-16]. Healbone bridges this gap by using wearable sensors and real-time feedback to mimic supervised guidance, addressing key limitations of traditional home-based methods (eg, poor adherence monitoring). This aligns with evidence that digital tools (eg, apps and wearables) improve exercise accuracy and compliance in chronic diseases. Exercise is known to reduce disease activity, pain, and stiffness in AS while enhancing physical function and quality of life [17]. However, its effectiveness depends on adherence, which is often low in home-based settings. Healbone tackles this.

Digital-powered remote system enables real-time quantitative assessment and precision-guided rehabilitation for patients with AS through sensor-based tracking, which can promote improvements in health and physical function. Studies indicate that the intensity, frequency, and duration of exercise intervention are crucial for its effectiveness [18]. However, previous studies have shown that the optimal mechanical load range for patients with spondyloarthritis is significantly narrower than that for normal individuals, proposing the concept of the “Goldilocks zone” for exercise load [19]. Supervised exercise rehabilitation guidance is constrained by time and location, which may affect long-term patient participation. In contrast, online exercise guidance platforms offer greater convenience and flexibility but may lack the ability to provide effective feedback, movement correction, and timely verbal encouragement [20]. Our system establishes a standardized exercise training model and provides real-time feedback and correction during exercise rehabilitation, effectively ensuring that patients engage in safe and effective exercise within their capabilities. This approach is expected to further enhance the clinical treatment outcomes for patients with AS and, expected to be widely implemented in outpatient rehabilitation settings.

This study has several limitations. One is that the patients are recruited for PLA General Hospital, which may limit the generalizability of the results to the rest of China and internationally. However, patients come from across the country, accordingly reducing some regional differences. Second, the differing monitoring strategies between groups-digital tracking for the intervention group versus self-reporting for the control group—may influence adherence comparisons. While this reflects real-world disparities between standard and technology-enhanced care, it introduces potential measurement bias. Future studies could use matched monitoring methods (eg, wearable devices for both groups) to isolate the effect of the digital platform itself, independent of adherence measurement tools. Finally, the exclusion of regularly active individuals may reduce the generalizability of our results. However, this approach aligns with prior trials aiming to evaluate novel interventions in controlled settings.

To our knowledge, this is the first randomized controlled trial to assess a digital functional exercise program, enhanced with real-time feedback and sensor-based monitoring, for improving disease activity and physical function in patients with AS. If effective, this program could offer a scalable, cost-efficient solution for home-based rehabilitation, addressing the critical need for precise, supervised exercise in AS management. By leveraging current technologies, this approach fills a significant gap in chronic disease care, with potential applicability to other musculoskeletal conditions.