The low mechanical loading of bones of wheelchair users due to nonambulation leads to low bone mineral density (BMD) [1,2]. Low BMD is problematic as it increases the risk of osteoporotic, or low-impact, fractures [2]. In ambulatory people, adding mechanical loading through physical activities, such as running, jumping, and resistance training, is a way of stimulating the bones to prevent bone loss [3-6]. Furthermore, resistance training and impact exercises are explicitly recommended to improve or at least maintain BMD, prevent falls, and thereby reduce the risk of osteoporotic fractures [6]. However, these recommendations are mainly based on studies investigating the effects of a whole-body training programs, which sometimes include jumping, weight-bearing activities, or a combination of resistance training and running [7], that may be inapplicable to nonambulatory wheelchair users.

Most research on bone health in wheelchair users has been performed on individuals with spinal cord injury (SCI). SCI is a risk factor for low BMD due to physiological changes such as reduced neuromuscular activity and changes to systemic hormones that create an unfavorable environment for bone formation, in addition to the acute offloading after injury and the reduction in mechanical stimuli in cases where the paralysis is chronic. Bone loss has been shown to range from 4% to 9% at the hip in the first 6 months after injury, stabilizing at a total loss of approximately 25% around 10 years after injury, when a new steady-state bone mass seems to be reached [8,9]. Throughout life, 25% of individuals with SCI experience at least 1 fracture, of which 70% occur due to a low-impact injury, such as moving from wheelchair to bed [10]. Consequently, interventions to improve BMD or at least slow bone loss are of high value.

In addition to reduced bone loading, low BMD can be related to suboptimal nutrition, such as chronic low energy intakes compared with the energy needs (low energy availability), low vitamin D blood levels, or low dietary calcium intakes [11-13]. In our previous study on Paralympic athletes [14], 25% displayed clinically low vitamin D blood levels (<50 nmol/L) and 75% displayed values below the International Olympic Committee (IOC) recommendations (<80 nmol/L) [12].

Physical activity and sports participation could be of great importance for the bone health of wheelchair users with low mechanical bone load. However, increasing general physical activity per se does not appear to solve the problem of low BMD in wheelchair users. This is evidenced by the finding that reduced bone health is more prevalent in sport-active individuals with a disability compared with the general population. Our group has recently found that 31% to 34% of Dutch and Norwegian Paralympic athletes display clinically low BMD (z score <–1.0) at the hip and femoral neck respectively [14]. The low BMD was most prominent in the wheelchair-using athletes, of whom 46% to 55% displayed clinically low BMD at the hip. In addition, 30% of wheelchair-using athletes displayed low BMD at the spine, compared to only 18% among the ambulatory Paralympic athletes [14]. Another study of Norwegian Paralympic athletes found similar results [15]. These findings suggest that the mechanical loading of the hip and spine achieved by physically active wheelchair users may not be adequate for maintaining bone health. In the adapted exercise literature, most interventions aiming to improve BMD are studies in patients with SCI, most commonly during the acute and subacute phases after injury [16]. Therapies such as passive standing [17,18], vibration therapy [19,20], and exercise with functional electrical stimuli [21-23] have shown little to no effect on BMD. However, the effects of high-load resistance training on bone health in wheelchair users are unknown. To our knowledge, no study has yet investigated an upper-body high-load resistance training program for improving the bone health of wheelchair users.

Alongside the effects of resistance training on bone health, several positive implications on physical health parameters can be expected, such as increased strength and improved body composition and blood health markers [24-28]. Furthermore, physical activity and participation in sport are major positive contributors to mental health, influencing well-being and quality of life [24]. Therefore, the primary objective of this randomized controlled trial (RCT) is to investigate the effects of upper-body high-load resistance training on the BMD of the lumbar spine and hip in wheelchair users. The secondary objectives are to investigate the effects of the intervention on (1) bone and physical health parameters, such as blood bone turnover markers, nutritional status, body composition, and maximal muscular strength, and (2) motivation for exercise and mental health indicators such as well-being, exhaustion, and vigor.

This protocol describes a multisite, single-blinded, 2-armed RCT comparing a 24-week resistance training intervention combined with nutritional optimization (training) to nutritional optimization only (control).

The study was conducted at 3 sites in Norway: Norwegian School of Sport Sciences (NIH, the main site), Western Norway University of Applied Sciences (HVL) in Bergen, and Norwegian University of Science and Technology (NTNU) in Trondheim.

Participants attended one of the 3 study sites for eligibility screening before being included in the RCT (T0). Those meeting the eligibility criteria were invited for baseline testing (T1), and thereafter intervention midpoint testing (T2, week 12 or 13 approximately), postintervention testing (T3, week 25), and a final follow-up testing (T4, 6-18 months later); overview in Figure 1. All intervention training was conducted at the study site, at a partnering commercial fitness centre local to the participant’s home. Within 6 months of completing the study, individual information about the development in BMD, body composition, maximal muscle strength, and nutritional status was provided to each participant in writing, with the option of over-the-phone counseling.

To be enrolled in the RCT, participants were screened for the following eligibility criteria: (1) BMD z score ≤0 SD at the spine, total hip, and femoral neck; (2) mainly nonambulant (using the wheelchair for ≥50% of the time); (3) aged 18 to 60 years; (4) ability to perform key exercises and maximal muscle tests; (5) nonprogressive impairment and, in the case of SCI, ≥2 years since the time of injury; (6) no comorbidities or use of medications that affect bone metabolism, nutritional uptake, or metabolism of vitamin D or calcium or that contraindicate high-load resistance training (eg, heart disorders); (7) not pregnant or menopausal; (8) no medical (eg, bisphosphonates) or physical (eg, functional electrical stimuli) treatment of low BMD; (9) no fractures affecting measured sites or contraindicating strength testing or training; and (10) no language or cognitive barriers interfering with the ability to understand study procedures or perform the training program. In addition, the ability to attend the laboratories and gym at the test site a minimum of 3 times over the study period was necessary for participation. A detailed list of eligibility criteria is provided in Multimedia Appendix 1. The inclusion of participants was authorized by the principal researcher in collaboration with senior researchers on the team and the study medical adviser.

Participants were recruited via Sunnaas Rehabilitation Hospital; the Norwegian Confederation of Sports; Sports cluster of Western Norway (Norwegian: Idrettsklynge Vest); and the Para Sport Centre in Trondheim, Norway, as well as advertisements in local media, the relevant national patient interest groups, and in social media channels. Participants expressed interest via a web-based registration form containing contact details and date of birth.

Power calculation was performed using G*Power (version 3.1; Heinrich-Heine University) [29]. The primary outcome in this study is the change in BMD of the spine. Means and SDs for a relevant difference between the training and control groups are based on 2 previous studies: one validation study in 345 healthy adults [30] and one 24-week bone-strengthening training intervention in a youth cohort with obesity [7]. Using the lowest significant increase in BMD of the spine of 3.5% and SDs from the 2 studies, 14 to 32 (average of 23) persons per group were estimated to achieve 80% power at a significance level of P≤.05. To account for a potential 25% dropout rate, the aim was to include 30 participants in each group. To achieve this, it was estimated that 120 to 150 participants would need to be screened. Of approximately 50,000 wheelchair users in Norway (according to the Norwegian Labour and Welfare Administration [31]), we estimated to have a pool of 1000 possible participants from the 3 main cities in Norway.

Eligible participants were randomly allocated to 1 of the 2 groups, the training (training and nutrition, n=30) or the control (nutrition only, n=30) group, directly after completion of baseline test procedures. The randomization list was created by a senior researcher not directly involved in the study procedures, using a web-based randomization tool [32]. The randomization list was stratified by physical activity status (active: high activity categorization in the International Physical Activity Questionnaire [33,34] or actively participating in sports for >6 months at least twice per week or nonactive) and sex (male or female), with a 1:1 allocation using random block sizes of 4, 6, and 8. To ensure that allocation concealment was sequentially numbered, nontransparent and sealed envelopes containing the treatment allocation information were prepared by a person not involved in the study. Blinding is upheld for the main outcome for the main researcher during analyses. The groups were checked for distribution of the number of participants with SCI, age, medication use, degree of injury or disability, and degree of wheelchair use.

The primary outcome of the study is the change in BMD of the lumbar spine, whereas secondary outcomes are changes in other markers of bone health, such as blood bone turnover markers; parameters of muscular, functional, and psychological dimensions; nutritional status; and other markers of overall health. All outcomes and the corresponding assessment methods are provided in Multimedia Appendix 3. Most assessments were conducted by the main researcher. Dual-energy x-ray absorptiometry (DXA) scans at the NIH and Bergen sites were performed by the main researcher, while the DXA scans at the NTNU site were performed by trained hospital staff. DXA scans were performed unblinded, while all DXA scan analyses were performed by the main researcher after blinding to group allocation. Trained bioengineers and researchers collected blood samples, while trained nutritionists or dieticians conducted the 24-hour dietary recalls.

This study protocol has received ethics approval from the Regional Committee for Medical and Health Research Ethics South-East, Norway (2023/458384). Protocol modifications have been approved in revised versions of ethics approval, and revisions have been communicated to all participants. Multimedia Appendix 4 provides the final approved protocol. In case of any serious adverse events, Norwegian health insurance applied to all participants, and any copayment was to be covered by the responsible institution (NIH).

All screened participants provided written informed consent before any procedures (including at the screening visit) and before being included in the RCT (Multimedia Appendix 5). Consent can be withdrawn at any time during the study period and until data have been published or have been made unidentifiable. The consent also covers secondary analyses without additional consent. All data were collected and deidentified with a study participant identification number. All participants received NOK 500 (approximately US $45-50) as compensation for participation. Furthermore, they obtained free access to a local fitness centre as part of the intervention period (training group). The control group obtained the opportunity to receive the training program and 4 weeks of supervision at the test site. After study completion, the participants with BMD z scores <–2.0 were advised to seek medical support. The study was preregistered in ClinicalTrials.gov (NCT05615402) on November 14, 2022, and in Open Science Framework [47] on January 4, 2023, and reporting follows the SPIRIT (Standard Protocol Items: Recommendations for Interventional Trials) guidelines (Multimedia Appendix 6).

A raw data copy was stored before data handling, including range checks and scatter plots to find potential outliers. Any corrections or deletions were saved in working file copies. Invalid data points were removed from the working files but not the raw files, for example, lumbar spine BMD values where the scan contains implants or metal artifacts. Questionnaire data were coded into appropriate categories and scored according to the questionnaire guidelines. Invalid questionnaire responses were sought to be collected again, if possible, either at a subsequent visit or by telephone interview. Data are shared between the collaborating institutions (HVL and NTNU) and the main site (NIH) via shared access to online storage (OneDrive for 365, Microsoft Corporation) for as long as practically necessary. Only key project team members have access to the data. The qualitative data are only available to 2 researchers at NIH.

BMD and participant characteristics at screening will be described and analyzed cross-sectionally, and linear multiple regression models will examine associations between BMD and other baseline data (such as age, sex, diagnosis category, and Spinal Cord Independence Measure mobility subscore) in the analyses of the RCT. The cross-sectional data will be analyzed descriptively, and differences between subgroups as well as explanatory factors of BMD will be analyzed with linear regression analyses. To analyze the group effects over the intervention period, a linear mixed model for repeated measures with an unstructured covariance structure will be used. Random intercept for participants and fixed effects of time, treatment, and their interaction, as well as the dependent variable value at baseline and study site, will be included in the model. The interaction term accommodates different patterns of change over time between intervention and control groups. The mixed model allows all outcome data to be used, regardless of whether an individual has complete data, making these models consistent with an intention-to-treat analysis. Correlation tests between the delta change of main parameters and background variables will also be performed. Data will be analyzed both per protocol and by intention to treat [48]. A P value <.05 will be used to indicate statistical significance. Data will be analyzed using SPSS (IBM Corp) and R studio (Posit PBC). The qualitative data will be analyzed using reflexive thematic analysis [49].

Between December 2022 and November 2023, a total of 104 interested wheelchair users were prescreened for eligibility before the screening visit (T0), of whom 72 (69.2%) were invited and 66 (63.5%) attended. See the participant flowchart in Figure 4. On the basis of the screening assessments, 52 (50%) eligible participants were invited to take part in the RCT, and ultimately 45 (NIH: n=33, 73%; HVL: n=5, 11%; NTNU: n=7, 15%) were included and underwent a baseline visit (T1). Following this, they were randomly allocated to 1 of the 2 groups: 24 (53%) to training and 21 (47%) to control. At the midpoint visit (T2), 36 (n=17, 47% and n=19, 53%, respectively) participants were reassessed and 33 (n=14, 42% and n=19, 58%, respectively) completed the study (T3). The final postintervention visit was completed in April 2024. Of the 22 participants who completed the study at the NIH test site, 15 (68%) revolunteered to attend a 6- to 18-month postintervention visit (T4) at the end of 2024.

Data analysis of data collected at the screening visit (T0) commenced in spring 2024, while analysis of data collected at the baseline and retest visits began in autumn 2024. The first results of this study are expected to be disseminated in peer-reviewed journals by the end of 2025.

The anticipated findings are that those randomized to the high-load resistance training program in combination with nutrition optimization will improve BMD at the spine and hip, and compared to a control group receiving nutrition optimization only.

In addition to the effects on the primary outcome, we further expect that the exercising group will increase both lean mass and muscular strength. Three moderate-high load sessions per week have been suggested to be effective in increasing lean mass and maximal strength [50,51]. As untrained individuals tend to have a greater relative increase in both outcomes, we expect a difference in effectiveness between the untrained and previously trained participants, and little to no effect in the nutrition optimization only group. Finally, we expect positive effects on the participants’ physical and mental health in general.

This study will give insight into the effects of a novel upper-body high-load resistance training program on BMD at the spine and hip, which we hypothesize will improve after intervention. While there is a knowledge gap regarding how to optimize bone health in wheelchair users nonpharmacologically, existing evidence in ambulatory people allows us to hypothesize that the combination of weight-bearing exercises and optimizing key nutrients for bone formation will be effective for the prevention and treatment of low BMD. Improvements of 2% to 4% over 12 to 24 weeks of resistance training have been shown in ambulatory individuals. In this study cohort of mainly nonambulatory wheelchair users, with far less or no stimuli from standing and walking, we hypothesize the same level of effect despite not including exercises typically performed in other studies (eg, squats or leg press and jumps). We further hypothesize that optimizing nutrition alone is not enough to accrue bone mass in this group, as purported by the mechanostat theory [52].

A total of 60 adult wheelchair users with nonprogressive impairment and high enough function to perform the exercises in the resistance training program were sought from both sports-active and nonactive environments. Owing to low recruitment in the early stages of the study, the NIH study site conducted several rounds of recruitment and enrollment over almost 12 months. However, we were unable to achieve the target numbers, which need to be accounted for in the dissemination of the results of the study. A more thorough process in estimating the feasibility of recruiting wheelchair users to such an extensive study protocol would have been valuable.

Trial results are provided in an individual report to all participants. The main findings will be published in several open-access publications, following Vancouver guidelines for authorship. We plan to grant public access to the full protocol, participant-level dataset, and statistical code. Furthermore, the results and experiences from the study will be communicated to the scientific community at international conferences and to health care professionals and other relevant groups in workshops and public presentations.

In conclusion, this is the first RCT to implement a long-term high-load resistance training and nutrition program focusing on improving bone health in wheelchair users. Furthermore, including a target group of both sport-active and nonactive participants allow us to compare bone health status at baseline and investigate the effects of the intervention in these subgroups. This study will add valuable scientific and practical information on the effects of targeted training and nutrition for bone health, as well as the coinciding influence on motivation for exercise, along with the general physical and mental health of wheelchair users.