NEOBIC-XR: a home-based extended reality exercise platform and multicenter evaluation framework for supportive oncology care

Exercise is increasingly recognized as part of supportive cancer care, but many patients face barriers to structured, safe, and sustainable exercise during treatment. Extended reality (XR) may support home-based delivery by combining immersive guidance, standardized sessions, and embodiment-related features. However, reproducible oncology deployment models remain limited. This Methods paper describes NEOBIC-XR, a home-based XR exercise platform for people with cancer, and the multicenter workflow used to evaluate its implementation. NEOBIC-XR integrates multilingual guidance, avatar-based self-representation, structured tutorials, adjustable intensity, mixed-reality components, and linked subjective, telemetry, and movement-derived data streams. The associated evaluation is a multicenter, implementation-oriented methods study conducted in three hospitals in Spain, with common baseline and 3-month follow-up assessments. Adults with stage IV solid tumors who are clinically suitable for light-to-moderate exercise are enrolled in the XR-supported pathway or the Written-recommendation pathway. XR-supported participants complete a 4-week home program with three prescribed sessions per week; Written-recommendation participants receive evidence-based physical activity recommendations and complete the same hospital-based questionnaires. The main planned comparative endpoint is baseline-to-3-month change in self-reported physical activity. Additional questionnaire-defined domains include health-related quality of life, symptom burden, psychological distress and screening items, cognitive complaints, supportive care needs, nutritional indicators, and other supportive-care measures. XR-specific analyses will describe first-to last-day changes in exercise-related readiness, coping self-efficacy, exercise self-efficacy, wellbeing, avatar representation, trainer-led versus body-double tutorial delivery, adherence, tolerability, embodiment-related experience, telemetry, and movement-derived indicators. Overall, this paper presents NEOBIC-XR as a privacy-preserving, analyzable, and clinically deployable methodological platform for exploring how home-based XR exercise can be implemented, experienced, and evaluated in oncology supportive care.

Cancer treatment affects many aspects of patients’ lives beyond disease status. Throughout the disease trajectory, many patients experience fatigue, pain, reduced physical capacity, loss of confidence in movement, and difficulty maintaining everyday activity. These problems may lead to physical deconditioning, functional decline, and poorer quality of life (Muhandiramge et al., 2022). Exercise is therefore increasingly recognized as a key component of supportive cancer care, not only to preserve physical function but also to reduce cancer-related fatigue and improve health-related quality of life (Campbell et al., 2019; Mustian et al., 2017; Schmitz et al., 2019). International guidelines support combined aerobic and resistance exercise across the cancer continuum and increasingly frame exercise as part of oncology care rather than as an optional lifestyle recommendation (Campbell et al., 2019; Schmitz et al., 2019).

Despite this, exercise is often not initiated, maintained, or progressed consistently in routine practice. Patients may face treatment-related symptoms, fluctuating energy levels, fear of overexertion, uncertainty about how to exercise safely, reduced exercise self-efficacy, and limited access to structured supervision (Finch and Benham, 2024). At the system level, exercise remains insufficiently integrated into routine oncology pathways, and referral, implementation, and resource barriers continue to limit its delivery (Kennedy et al., 2022; Ng et al., 2021; Kampshoff et al., 2014). Thus, an important challenge in exercise oncology is not only to demonstrate benefit, but also to develop interventions that patients can realistically access, understand, tolerate, and sustain.

Home-based digital interventions may help address some of these barriers by improving accessibility and flexibility (Batalik et al., 2024). Within this context, virtual reality (VR) and broader extended reality (XR) approaches are of particular interest because they can provide exercise guidance in immersive, standardized, and interactive environments (Buche et al., 2022; Levin et al., 2015; Guy et al., 2023). Recent reviews suggest that VR-based rehabilitation in cancer may be feasible, safe, and beneficial in selected settings, although the evidence remains heterogeneous and is still concentrated in a limited number of populations and use cases, particularly upper-limb rehabilitation after breast cancer treatment (Su et al., 2024; Eid et al., 2025; Zhou et al., 2021; Feyzioğlu et al., 2020).

Importantly, XR exercise is not a single intervention category, and not all systems address the same patient needs. Beyond presenting movements within a virtual environment, full-body-tracking approaches may support exercise through a visible virtual self-representation, creating conditions for embodiment, including body ownership, agency, and self-location (Su et al., 2024; Guy et al., 2023). This may be particularly relevant in oncology, where symptom burden and physical deconditioning often coexist with reduced confidence in movement and difficulty sustaining exercise routines. In addition, structured digital guidance, adaptive pacing, and tolerable home-based delivery may all be important for long-term adherence.

A relevant gap therefore remains between the general recommendation to promote exercise in oncology and the availability of home-based interventions that are both clinically usable and methodologically reproducible. Although XR may offer advantages over conventional digital formats, evidence is still limited regarding how such systems can be deployed in patients receiving cancer treatment, how acceptable and tolerable they are in repeated home use, and whether design features related to embodiment, guidance, and adaptive delivery may contribute to exercise support in this setting. Addressing these questions is important for the development of scalable and evaluable XR-based supportive care strategies in oncology.

Accordingly, this paper has a double aim. First, it describes the NEOBIC-XR platform, including its hardware and software configuration, virtual environments, avatar and tutorial structure, home-deployment workflow, and privacy-preserving data architecture. Second, it presents the evaluation framework used to study the platform and selected design dimensions under real home-use conditions. The main planned pathway comparison focuses on baseline-to-3-month change in self-reported physical activity, together with repeated patient-reported outcomes available in the common baseline and 3-month follow-up assessment packets, between the XR-supported pathway and the Written-recommendation pathway. Additional XR-specific first-day/last-day measures, session-level measures, and design-related variables are examined as proximal outcomes and initial design and implementation analyses within the XR-supported pathway. Thus, the manuscript presents a reproducible Methods paper describing the platform, its deployment procedure, and the evaluation framework used to study it in oncology supportive care.

Platform hardware, software, and virtual environments

NEOBIC-XR was developed in Unity 6000.3.1f1 and deployed on Meta Quest 3 standalone head-mounted displays. The main software stack combined Unity XR Interaction Toolkit (version 3.3.1), XR Plugin Management (version 4.5.3), OpenXR Plugin (version 1.16.0), Timeline (version 1.8.10), and Localization (version 1.5.9), together with Meta XR Core SDK (version 81.0.0), Meta XR Interaction SDK (version 81.0.0), Meta XR Movement SDK (version 77.0.1), and the OVRLipSync Unity Plugin (version 1.54.0). The software stack also included Unity Addressables (version 2.7.4). These components supported headset interaction, hand tracking, localization, pre-exercise tutorial sequences, mixed-reality transitions, full-body avatar retargeting, trainer-avatar lip synchronization, and session-level movement recording derived from the retargeted avatar rig. Voice content for the virtual trainers was generated with ElevenLabs (version 3) and integrated as pre-produced audio assets. Archived trainer scripts, voice assignments, and audio-line inventories are provided in Supplementary Data Sheet S1.

The required materials for reproducing the workflow are: Meta Quest 3 headsets; the NEOBIC-XR Unity application and associated XR packages listed above; study-managed headset accounts and restricted deployment access; pre-produced multilingual trainer audio; avatar and tutorial assets; optional participant-specific avatar-generation access; the written exercise-recommendation sheet; baseline, follow-up, XR-specific, post-session, onboarding, and telemetry documentation; and a center-managed procedure for extracting and linking local session logs and movement-derived files.

The intervention uses three main virtual environments distributed across the 4-week program, together with mixed-reality tutorial components. Week 1 takes place in a Roman Pantheon environment with a trainer adapted to that setting. Week 2 uses a green-field environment, and Week 3 a beach environment, each with a trainer matched to the corresponding scene. Week 4 reuses the environments and weekly structures from the preceding weeks. Representative examples of these environments, embodiment views, and trainer-guided tutorial contexts are shown in Figure 1.

Animation assets, tutorial materials, and movement production

The animation pipeline combined several sources. Boxing and squat demonstrations were adapted from Mixamo animations, the reaching demonstration was processed using DeepMotion, and the lateral lunge sequence was produced using camera-based motion capture processed through Rokoko Vision. Additional lower-body exercise animations were adapted from the Power Gym Training Animations package. These materials were integrated into exercise-specific tutorial sequences and adapted to the multilingual and round-based structure of the application.

All avatar representations used in NEOBIC-XR, including the standard avatar library and optional participant-specific photorealistic avatars, were integrated as Unity-compatible rigged avatar assets and configured for full-body retargeting using the Meta XR Movement SDK. Avatar assets were managed through Unity Addressables and instantiated at runtime according to the participant’s selected representation. During active XR use, the participant’s tracked movements were retargeted in real time to the selected avatar, allowing the virtual body to reproduce the participant’s movements. The selected avatar representation was also used as the visual body-double when required by the assigned tutorial condition.

Using the same retargeted avatar rig, the system records participant movement in derived kinematic form during use. Rather than storing raw video or image data, the application captures the positions and rotations of the retargeted avatar joints over time, enabling characterization of movement timing, execution trajectories, and exercise-related motion at the level of the embodied rig.

These pose files are treated as sensitive pseudonymized research data and are not interpreted as validated biomechanical, force, energy-expenditure, or clinical movement-quality measures unless supported by a separate validation analysis.

Operational workflow and XR intervention

The application follows a fixed macro-structure. Participants first select the interface language, then create a new local profile or continue an existing one, confirm their assigned avatar representation, and select the corresponding routine day. This is followed by the embodiment and warm-up phase, the exercise block, and the final breathing routine. After completion of the daily program, a replay mode is offered, allowing participants to repeat previously completed exercises voluntarily. Analytically, replay use is not considered part of the prescribed dose, but rather an indicator of continued willingness to exercise and preference for specific exercises.

Each prescribed daily routine includes four active exercises followed by the final breathing routine, and the weekly structure is repeated across the three scheduled sessions of that week. The exercise set combines upper-body target-based tasks, such as boxing and reaching, lower-body tasks, such as squats and lateral movements, and a guided mixed-reality lower-body module. The in-application labels low, medium, and high refer to relative task cadence, movement amplitude, or target demands within the program and should not be interpreted as standardized physiological-intensity categories. Participants are instructed to remain within the clinician-approved light-to-moderate range, to reduce the task level when symptoms increase, and to stop according to the safety rules described below. After avatar confirmation, the application initializes two parallel digital recording layers: an event-based analytics recorder, which captures session structure, stage transitions, routine selection, exercise progression, intensity choices, tutorial skipping, timing, scores, and replay behavior; and a pose recorder, which captures the joint trajectories of the retargeted avatar rig throughout use.

At the exercise level, the workflow follows a consistent structure. Each exercise begins with a pre-exercise phase composed of a tutorial followed by intensity selection. The tutorial contains two demonstrations, each with three short segments that show the exercise at low, medium, and high intensity. The first demonstration is always trainer-led. Before the second demonstration, a visible duplicate of the participant’s selected avatar becomes observable in the scene as a virtual body-double. In one condition, the trainer performs the second demonstration while the body-double remains visible in an observer position. In the other, the body-double performs the demonstration while the trainer remains visible and continues delivering the same instructional audio. This design preserves a common instructional framework while varying the visual source of the second demonstration. The two tutorial-format conditions are illustrated in Figure 2.

Tutorials are not always skippable. They are preserved when an exercise is experienced as novel, such as during its first presentation within the weekly schedule or in the Week 4 routines, as participants have not been exposed to that specific exercise in the last 7 days. By contrast, tutorials may be skipped in repeated sessions when the same exercise has already been practiced recently. This rule was intended to balance procedural consistency with ecological usability during repeated home use. After the tutorial, participants select the exercise intensity for the upcoming round. The exercise then proceeds through repeated rounds separated by rest periods. Once an exercise is completed, a continuation panel appears before the next exercise begins. After the final scheduled exercise, the session ends with the breathing routine.

Deployment workflow and privacy-preserving implementation

A major implementation challenge of the study was to support multicenter home deployment while preserving participant privacy and maintaining a workflow that remained feasible across the three participating hospitals. The deployment model was therefore designed to be center-managed and user-centered: headsets were prepared, updated, and distributed through the study workflow so that participants received a ready-to-use device and were not required to manage installation procedures or commercial account settings themselves.

The application was distributed through a restricted Meta Horizon Store deployment rather than as a public consumer release. Access was limited to study-managed headset users created specifically for the devices used in the study. This allowed controlled access to study builds, enabled application updates during the study when required, and supported center-side integration of participant-specific photorealistic avatars into updated builds before a headset was delivered or re-prepared.

For participants who opted for the photorealistic-avatar workflow, a single frontal facial photograph is used as input to the web-based AvatarSDK MetaPerson Creator. The generated avatar is downloaded as an FBX file with a 1024 x 1024 texture resolution and imported into the NEOBIC-XR Unity project for integration into the participant-specific study build. The corresponding avatar was not openly visible in the avatar-selection menu by default. Instead, participant-specific avatars remained locked until the appropriate unlock code was entered. This access-control mechanism reduced the possibility that another participant or staff member handling the same device could browse or accidently select an avatar that did not belong to them. Participants who did not opt for this workflow continued using the standard avatar library.

An additional privacy-preserving measure concerned headset accounts. Participants were not required to create or manage personal Meta accounts as part of study participation. Devices were provisioned through the study workflow and delivered in a ready-to-use state. At the application level, progress and digital records were organized around local profiles rather than personal names. Persistent profile data, structured session logs, and movement-derived pose files were stored locally and subsequently linked to the study database through the local study workflow rather than through direct identifying information. Because movement files contained derived kinematic trajectories rather than raw image or video streams, the digital capture layer remained compatible with a privacy-preserving model for clinical home use.

Methods

Study design, allocation, and objectives

This is a prospective, multicenter, non-randomized implementation and methods study conducted in three Spanish hospitals, with a common 3-month follow-up after baseline. The design has two linked components: a planned pathway comparison between the XR-supported pathway and the Written-recommendation pathway, and a nested methodological evaluation within the XR-supported pathway.

At the main comparative level, eligible participants who accept XR-supported home use and can be accommodated within headset and onboarding capacity enter the XR-supported pathway. Participants in the XR-supported pathway receive the 4-week NEOBIC-XR home intervention, with three prescribed sessions per week. Participants in the Written-recommendation pathway receive written evidence-based physical activity recommendations and complete the same common hospital-completed baseline and 3-month follow-up questionnaires, but do not receive the headset, application, telemetry, or pose-recording components. The intended recruitment structure is approximately one participant in the XR-supported pathway for every three participants in the Written-recommendation pathway. This structured allocation reflects device availability, supervised onboarding capacity, and participant acceptance of home XR use; it is prespecified but not randomized.

Both pathways contribute to the common baseline-to-3-month comparison. Measures collected only in the XR-supported pathway, including first-day/last-day questionnaires, post-session questionnaires, telemetry, and movement-derived pose data, are analyzed as proximal, design, and implementation data rather than as common comparative outcomes.

Within the XR-supported pathway, two intervention dimensions are examined. Avatar representation is studied as a participant-level observational factor, comparing default-avatar use with optional participant-specific photorealistic self-resembling avatar use under acceptance-based uptake. Tutorial format is studied as a repeated day-level factor. At creation of a new participant profile, a computer-generated 1:1 random assignment places the participant in one of two counterbalanced order groups. One group begins with the body-double condition and the other with the trainer-led condition. The assigned condition applies to all tutorials presented within a given prescribed routine day and alternates on the subsequent prescribed routine day; the assigned order is retained across the intervention. This randomization concerns only tutorial-order assignment; allocation to the XR-supported and Written-recommendation pathways remains non-randomized.

The main planned comparative objective is to compare baseline-to-3-month change in self-reported physical activity between pathways. Additional comparative objectives concern other patient-reported outcomes repeated in comparable form at baseline and 3 months. XR-specific proximal objectives examine first-day to last-day change within the XR-supported pathway, and initial design and implementation analyses characterize avatar representation, tutorial format, adherence, completion, tolerability, embodiment-related experience, telemetry-derived indicators, and movement-derived indicators of actual use. The overall assessment flow and nested XR-supported components are summarized in Figure 3.

Study setting and participants

The study is conducted in three hospitals in Spain: Hospital Universitario Central de Asturias (Oviedo), Hospital General Universitario de Ciudad Real (Ciudad Real), and Hospital Universitario Infanta Leonor (Madrid). All participants complete the baseline oncological assessment and the main baseline questionnaire at the recruiting hospital before starting any home-based study component. The common 3-month follow-up questionnaire is also completed at the hospital.

Participants enrolled to the XR-supported pathway undergo a supervised hospital-based onboarding session and, once safe and functional use of the system has been confirmed, receive the headset for home use during the 4-week intervention period. The intervention is therefore delivered at home in the XR-supported pathway, whereas participants in the Written-recommendation pathway follow a non-headset pathway based on written evidence-based physical activity guidance and the same broader follow-up schedule.

To support consistent multicenter implementation, all centers follow a common workflow for device preparation, onboarding, and study management. Headsets are prepared and managed at the center level and provided to participants in a ready-to-use format for home deployment. Participants in the Written-recommendation pathway do not receive the application or headset and are followed according to the planned assessment schedule without home XR use. The written recommendation sheet used for the Written-recommendation pathway is provided in Supplementary Data Sheet S2.

The participant-facing onboarding guide used to support home deployment for the XR-supported pathway is provided in Supplementary Data Sheet S3.

Eligibility criteria

Eligible participants are adults with histologically confirmed stage IV solid tumors, irrespective of primary tumor origin, who are considered clinically suitable for light-to-moderate exercise and able to follow the study procedures at home. Inclusion requires an Eastern Cooperative Oncology Group performance status of 0 or 1, an estimated survival of at least 12 months at the time of stage IV diagnosis, initiation of first-line systemic anticancer treatment at study entry, and absence of comorbidities that would contraindicate physical exercise according to clinical judgment.

Participants are excluded if they have contraindications to, or a high safety risk for, physical exercise. These include unstable or high-risk cardiac disease, such as unstable angina or uncontrolled ischemia, decompensated heart failure, uncontrolled arrhythmias or recent syncope, uncontrolled hypertension, severe symptomatic valvular disease, or a recent acute cardiovascular event without medical clearance for exercise. Exclusion criteria also include unstable respiratory disease, such as uncontrolled respiratory failure or resting hypoxemia, recent exacerbation of chronic obstructive pulmonary disease or asthma, or uncontrolled limiting dyspnea; acute infection, fever, or intercurrent clinical decompensation discouraging exercise; neurological, musculoskeletal, or balance disorders that compromise safety because of a high risk of falls or injury, or uncontrolled pain preventing activity; and oncological situations associated with significant mechanical risk that contraindicate unsupervised exercise, including bone metastases with high fracture risk, suspected or confirmed spinal cord compression or vertebral instability, or central nervous system lesions with unstable neurological manifestations, unless specific authorization and an adapted plan are in place.

Participants are also excluded if they are unable to understand or follow study instructions adequately. This includes a relevant language barrier without appropriate support, moderate-to-severe cognitive impairment, delirium, or decompensated psychiatric disease compromising adherence or safety, as well as sensory impairment preventing adequate use of study materials, such as non-correctable moderate-to-severe visual loss or severe uncompensated hearing impairment when audio-dependent material is required.

Finally, participants are excluded if they have limitations that preclude safe or feasible use of the study devices. These include essential tremor or other motor disorders substantially limiting safe handling of the controller or device, inability to use the required digital format in the absence of caregiver support, and intolerance or conditions making virtual reality inadvisable for safety reasons, such as severe motion sickness or vertigo, photosensitive epilepsy, or uncontrolled seizures.

Supplementary Data Sheet S4 is the shared NEOBIC clinician baseline and screening form used across the broader NEOBIC study workflow. For the present NEOBIC-XR Methods study, eligibility for the XR-supported and Written-recommendation pathways is defined by the XR-specific criteria described in this section; additional variables in the shared screening form are recorded for clinical characterization, safety screening, or broader NEOBIC documentation and do not expand eligibility for the present analysis.

Recruitment, consent, and enrollment

Participants are recruited through the collaborating oncology hospitals and complete study entry procedures at the recruiting center before starting any home-based study component. Enrollment begins with clinical screening and baseline oncological assessment, after which eligible patients are informed about the two study pathways and complete the main baseline questionnaire at the hospital. Potential participants are identified by the treating oncologists during routine clinical practice. All eligible patients receive oral and written information about the study, and written informed consent is obtained before any study-specific procedures are performed.

Allocation follows a predefined operational procedure at each center. On one fixed recruitment day per week, eligible patients may be offered the XR-supported pathway. Those who accept XR participation and can be accommodated within the daily onboarding and headset capacity are enrolled to the XR-supported pathway, with a maximum of two participants per day according to order of arrival to the oncology consultation. Eligible patients recruited outside the XR-capacity window, those who cannot be accommodated within the available headset/onboarding capacity, and those who do not accept home XR use enter the Written-recommendation pathway and receive written exercise recommendations. Allocation is therefore structured and prespecified, but not randomized; it is driven by operational capacity and participant acceptance rather than by random sequence generation.

Participants enrolled to the XR-supported pathway are scheduled for a dedicated study visit within 72 h. At that visit, they receive detailed information about the XR-supported pathway, undergo the onboarding procedure, receive the headset, and are provided with the XR-specific first-day, last-day, and daily routine questionnaires, which they complete at home during the XR-supported pathway. In addition, participants receive telephone and email contact details for support during the intervention period.

Participants enrolled to the Written-recommendation pathway do not undergo the XR onboarding visit. Instead, they receive the written exercise recommendations at the oncology department, while the main baseline questionnaire is completed at the hospital as part of the common study-entry workflow.

Within the XR-supported pathway, participant-specific photorealistic avatar generation is offered as an optional additional component. This procedure requires separate written informed consent and is performed only in participants who agree to it. Participants may withdraw consent for the XR-supported pathway or for the avatar-generation procedure at any time without leaving the broader study or compromising the planned follow-up in the corresponding pathway.

Ethics

The study will be conducted in accordance with the principles of the Declaration of Helsinki and the applicable national and institutional regulations for research involving human participants. The study protocol (NEOBIC-2025-01) was approved by the Research Ethics Committee with Medicines (CEIm) of the Principality of Asturias (approval number 2025.180; IRB00003099). The present manuscript corresponds to protocol version 1.0. Participant recruitment began in 2026 following completion of the required regulatory and site-initiation procedures and will continue until the target sample size is achieved. The anticipated end of recruitment is expected in 2027. Any amendment affecting eligibility criteria, study outcomes, intervention content, safety monitoring procedures, data management, or statistical analyses will be submitted for ethical review and approval as required, documented in the study records, and reported in all relevant publications and final study reports.

All participants provide written informed consent prior to the initiation of any study-specific procedures. Within the XR-supported pathway, the generation of a participant-specific photorealistic avatar is offered as an optional component and requires separate additional written informed consent.

Participant confidentiality is ensured through the use of coded identifiers and the separation of personal data from clinical and research datasets. Questionnaire data, local headset profiles, device-derived telemetry, and movement-derived pose data are handled using procedures that restrict access to authorized study personnel and prevent direct identification of participants. When applicable, photographs used for optional avatar generation are processed within a controlled workflow and are not linked to identifiable personal information in the study dataset. All data are stored and managed in accordance with applicable data protection regulations.

Safety, adverse events, and discontinuation

As the intervention combines exercise participation with headset-based home use, safety monitoring addresses both exercise-related events and XR-related tolerability. During onboarding, participants are instructed on how to interrupt a session, remove the headset, and stop the activity if symptoms occur.

Safety and tolerability during XR sessions are assessed using a post-exercise questionnaire completed after each session, provided in Supplementary Data Sheet S5. This instrument records session duration, interruption status, and reasons for non-completion when applicable. It also includes a safety section in which participants rate symptoms such as nausea, dizziness, headache, and muscle pain experienced during or immediately after the session.

These data provide a standardized measure of acute tolerability at the session level. Relevant safety signals include participant-reported symptoms and any interrupted or incomplete sessions. Reports of clinically relevant discomfort, symptom exacerbation, or session interruption are reviewed by the research team and may lead to continuation as planned, adaptation of the exercise protocol, temporary pause, or discontinuation of the XR-supported pathway.

For the purposes of safety monitoring, a serious adverse event is defined as any untoward medical occurrence that results in death, is life-threatening, requires inpatient hospitalization or prolongation of existing hospitalization, results in persistent or clinically significant disability or incapacity, or is judged by the investigator to represent an important medical event. Serious related harm is defined as a serious adverse event judged by the principal investigator or delegated clinical investigator to be possibly, probably, or definitely related to the XR-supported exercise intervention, headset use, onboarding procedure, or study-specific home-use procedure. Events reported by participants, caregivers, or clinical staff are reviewed by the local clinical research team, with relatedness and severity assessed by the responsible investigator in consultation with the treating oncology team when needed. Serious adverse events judged related to the intervention or study procedures are documented in the study records, reported according to institutional and ethics-committee requirements, and trigger clinical review before any decision about continuation, adaptation, temporary suspension, or discontinuation of the XR-supported pathway.

Repeated post-session assessments allow characterization of tolerability over time and will be used both descriptively and, where appropriate, analytically to examine whether acute symptoms or session interruption are associated with reduced adherence, lower session completion, or discontinuation.

Because participants begin active systemic anticancer treatment at study entry, they also undergo regular oncology follow-up during the intervention. For most treatment schedules, clinical visits occur every 7–14 days. For participants receiving treatment at 21-day intervals, an additional telephone contact is conducted at approximately day 14. These contacts provide opportunities to assess clinical status and report difficulties or adverse events potentially related to the XR intervention, in addition to session-level symptom recording and the on-demand telephone and email support pathway.

Sample size

The planned sample is determined by recruitment capacity across the participating hospitals and by the feasibility of implementing the XR-supported intervention at home. The main comparative assessment focuses on baseline-to-3-month change in total activity MET-min/week, calculated from the physical-exercise section included in the common baseline and 3-month follow-up patient questionnaire packets. Additional repeated patient-reported outcomes from the common baseline and 3-month questionnaires will be analyzed as secondary planned comparative endpoints. Within the XR-supported pathway, first-day to last-day changes in Godin physical activity, readiness to increase exercise, selected Cancer Behavior Inventory coping self-efficacy items, exercise self-efficacy, and wellbeing will be analyzed as XR-specific proximal outcomes.

The XR-supported pathway is planned for up to 40 participants. This number reflects headset availability, supervised onboarding capacity, home-deployment capacity, and the expected number of eligible participants who will be willing and able to use XR at home during cancer treatment. Recruitment is dynamic at the center level: eligible and consenting participants are enrolled to the XR-supported pathway when a headset and onboarding slot are available, whereas other eligible participants enter the Written-recommendation pathway according to the structured allocation workflow. Because headset availability, onboarding capacity, and participant acceptance limit access to the XR-supported pathway, recruitment is expected to yield approximately three participants in the Written-recommendation pathway for every one participant in the XR-supported pathway. If 40 participants are enrolled in the XR-supported pathway, the expected Written-recommendation pathway sample is approximately 120 participants, giving a total sample of approximately 160 participants.

Recruitment will end when the planned XR-supported pathway sample has been reached. Eligible participants already enrolled in the Written-recommendation pathway will be retained in the final sample and included in the planned analyses. Within the XR-supported pathway, initial analyses of avatar representation, tutorial format, telemetry-derived process variables, and movement-derived variables will be conducted within the planned XR-supported pathway sample and will estimate condition-specific responses and uncertainty for the platform’s design dimensions, especially the avatar-representation and body-double tutorial-format contrasts.

Timing, procedural checkpoints, and implementation contingencies

At the macro level, the study workflow comprises baseline clinical assessment, a 4-week home intervention period in the XR-supported pathway with three prescribed sessions per week, and a 3-month follow-up shared by both study pathways. At the session level, the sequence is fixed and includes language selection, local profile creation or continuation, avatar confirmation, routine-day selection, embodiment/warm-up, exercise performance, and the final breathing routine. At the exercise level, the sequence consists of a pre-exercise tutorial, intensity selection, repeated exercise and rest cycles, and transition to the next exercise or to session completion. This operational structure is aligned with the standardized clinician-facing onboarding workflow described in Supplementary Data Sheet S6.

The timing structure of the intervention is as follows: the baseline onboarding visit is criterion-based rather than fixed in duration and continues until the participant can complete the supervised checkpoints required for home use; the home intervention period lasts 4 weeks; the XR-supported pathway completes three prescribed sessions per week; each session includes a 5-min embodiment/warm-up phase, four exercises, one tutorial before each exercise lasting approximately one–1.5 min when fully played, six rounds per exercise, 30 s of active exercise per round, 10 s of inter-round rest, and a final 5-min breathing routine. In addition, session duration includes user-dependent interaction time related to language selection, profile confirmation, avatar confirmation, routine access, intensity selection, and between-exercise transitions. The expected base duration of one session is approximately 29–31 min, excluding this additional interaction-dependent time. Follow-up assessment is performed 3 months after baseline.

The baseline onboarding visit is intentionally defined as a supervised criterion-based procedure rather than a fixed-time session. Before home use is authorized, participants must demonstrate adequate headset fitting and comfort, correct Guardian definition and understanding of safe-area boundaries, basic controller and hand-tracking interaction, correct recentering, successful application launch, in-app interaction with the virtual buttons, avatar confirmation, access to the prescribed routine, and completion of the initial warm-up preview under clinician or study-assistant supervision. If one or more checkpoints are not successfully completed at the first attempt, re-training is provided during the same visit before home use is authorized.

To reinforce the supervised onboarding procedure during home use, participants are also provided with continued access to five short instructional videos: a brief NEOBIC-XR demonstration; headset fitting, safety, button, and charging instructions; basic controller-based interactions, including interface navigation, passthrough, boundary management, and volume adjustment; equivalent interaction guidance using hand tracking; and a first-steps tutorial showing how to locate and launch the application and complete the initial language, avatar, and routine selections. These materials remain available throughout the home intervention and complement telephone and email support from the study team.

A specific implementation contingency concerns the optional photorealistic-avatar pathway. When this option is used, avatar acquisition and deployment are completed at the center with the supervising clinician rather than initiated remotely at home. Delayed availability of a photorealistic avatar is therefore handled within the supervised center-based workflow and is not considered a home-side troubleshooting event.

During home use, common contingencies include cybersickness symptoms, session interruption, inability to complete a prescribed routine, tracking problems, Guardian redefinition, or other headset-related difficulties. Participants are instructed to stop the session if relevant symptoms or technical difficulties occur and to contact the study team using the support pathway provided at onboarding. Depending on the nature and persistence of the problem, the operational response may include remote guidance, repeat onboarding at the center, temporary suspension of the home intervention, or discontinuation from the XR-supported pathway when safe or feasible continuation cannot be ensured. Participants who discontinue XR use but remain enrolled in the broader study will continue the planned follow-up and will be retained in pathway-level analyses; discontinuation and non-adherence will additionally be recorded as implementation and feasibility outcomes.

Outcome measures and assessment schedule

Outcome collection is organized to avoid treating all variables as if they were measured in the same way. Three levels are distinguished: common baseline-to-3-month measures collected in both pathways, proximal first-day/last-day measures collected only in the XR-supported pathway, and session-level XR-specific measures linked to telemetry and movement-derived pose data.

Table 1 summarizes the assessment schedule and clarifies which measures are common to both pathways and which are specific to the XR-supported pathway.

At the common comparative-assessment level, all participants complete the patient baseline questionnaire and the patient 3-month follow-up questionnaire at the hospital. The clinician baseline form is provided in Supplementary Data Sheet S4, the patient baseline questionnaire in Supplementary Data Sheet S7, and the patient 3-month follow-up questionnaire in Supplementary Data Sheet S8. The baseline packet covers sociodemographic information, treatment expectations, well-being and screening items, health-related quality of life, coping, symptom distress, concentration and cognitive complaints, self-reported physical exercise, supportive care needs, patient participation, acceptance of illness, social network, nutritional status, diet, fatigue, and spirituality. The 3-month packet repeats the constructs available for longitudinal comparison and adds follow-up adverse-effect items. Accordingly, controlled longitudinal analyses will be restricted to measures collected in comparable form at both time points; baseline-only and follow-up-only variables will be summarized descriptively or used as contextual variables where appropriate.

Within the XR-supported pathway, participants also complete the first-day and last-day assessments at home across the 4-week intervention period (Supplementary Data Sheet S9). This instrument includes the Godin Leisure-Time Exercise Questionnaire, a readiness ruler-type item, selected Cancer Behavior Inventory items, the Self-Efficacy for Exercise Scale, and brief thermometer-style ratings of sleep, fatigue, pain, stress, overall well-being, and perceived medical care. These measures provide a proximal pre/post description of the XR-supported phase but are not common comparative outcomes.

After each prescribed XR session, participants complete a daily routine/post-session questionnaire at home (Supplementary Data Sheet S5) assessing presence, body ownership, agency, tolerability symptoms such as nausea, dizziness, headache, and muscle pain, perceived exertion, and, where recorded, interruption and duration fields. These subjective session-level data are linked with two device-local streams described in Supplementary Data Sheet S10: structured telemetry and movement-derived pose data from the retargeted avatar rig. Telemetry records profile and session identifiers, language, avatar condition, routine and day selection, timestamps, exercise progression, tutorial events, intensity choices, rest periods, skip events, completion status, replay events, and score-related variables.

This linkage allows session-level questionnaires, event-based behavior, and movement-derived data to be analyzed in relation to the nested XR factors. In particular, avatar representation and tutorial-order assignment are stored at participant/profile level, whereas the tutorial condition for each prescribed routine day is determined by the assigned counterbalanced order and routine day. Selected intensity and other exercise-specific behavioral variables are recorded at exercise level. Selected intensity is therefore treated as an exercise-level behavioral indicator for the initial tutorial-format analyses, alongside completion, interruption, replay behavior, and other use indicators.

Common comparative, XR-specific proximal, and design outcomes

The outcome hierarchy distinguishes planned comparative endpoints, XR-specific proximal outcomes, and initial XR-specific design and implementation analyses.

The primary planned comparative endpoint is baseline-to-3-month change in self-reported physical activity, calculated from the physical-exercise section administered in the common baseline and 3-month patient questionnaire packets. At each time point, walking, moderate activity, and vigorous activity will be scored using the reported number of minutes per day and number of days per week. Values marked as “+180” minutes will be coded as 180 min per day. Total activity will be expressed as MET-minutes per week using an IPAQ-style scoring approach: walking MET-min/week = 3.3 × walking minutes/day × walking days/week; moderate activity MET-min/week = 4.0 × moderate-activity minutes/day × moderate-activity days/week; vigorous activity MET-min/week = 8.0 × vigorous-activity minutes/day × vigorous-activity days/week. The primary physical-activity score will be the sum of walking, moderate, and vigorous MET-min/week. Sitting time and days spent barely moving will not be included in the primary activity score and will be analyzed descriptively as sedentary-behavior indicators. The primary comparative outcome will be the change in total activity MET-min/week from baseline to 3 months.

Secondary planned comparative endpoints are baseline-to-3-month changes in repeated patient-reported outcomes collected in comparable form in both common questionnaires, including health-related quality of life, symptom burden, psychological distress and screening items, concentration and cognitive complaints, supportive care needs, and nutritional indicators. Variables collected only at baseline, including sociodemographic and contextual psychosocial items, will be summarized descriptively or considered as covariates. Follow-up-only adverse-effect items will be treated as contextual, safety, or tolerability-related descriptors.

XR-specific proximal outcomes are first-day to last-day changes within the XR-supported pathway in Godin physical activity, readiness to increase exercise, selected Cancer Behavior Inventory items, exercise self-efficacy, and brief symptom and wellbeing ratings. These are interpreted separately from the comparative assessment because they are collected only in the XR-supported pathway.

Initial XR-specific design and implementation analyses address avatar representation, tutorial format, post-session embodiment and tolerability ratings, perceived exertion, prescribed-session completion, replay use, telemetry-derived session behavior, and movement-derived variables from the retargeted avatar rig. These analyses characterize usability, embodiment-related experience, adherence, movement execution, and preliminary signals that can guide future study design.

Statistical analysis plan

The main comparative analysis will compare change from baseline to 3-month follow-up between the XR-supported pathway and the Written-recommendation pathway for the primary planned comparative endpoint: total activity MET-min/week calculated from the common patient questionnaire packets as defined above. Because allocation is structured, non-randomized, and partly acceptance-based, between-pathway comparisons will be interpreted as adjusted between-pathway comparisons within the implementation context.

Baseline characteristics will be summarized descriptively and compared between pathways to assess initial comparability. The primary analysis will use a linear mixed-effects model including time, pathway, and their interaction as fixed effects, participant as a random effect, and study center as a fixed effect. The pathway-by-time interaction will estimate the between-pathway difference in change over time. Prespecified covariates will include age and sex. The baseline value of the outcome is represented within the repeated-measures structure; baseline-adjusted follow-up models may be used as sensitivity analyses when appropriate.

To assess the robustness of the between-pathway comparison to the non-randomized allocation procedure, an additional sensitivity analysis will address potential baseline imbalance between the XR-supported and Written-recommendation pathways. The primary model will be repeated with additional adjustment for clinically relevant pre-allocation variables available in both pathways, including symptom burden and selected clinical and treatment-related characteristics, subject to the final sample size and data distribution. This analysis will evaluate whether the estimated between-pathway difference remains consistent after accounting for measured baseline differences.

Additional comparative analyses will apply analogous models to patient-reported outcomes available in comparable form at baseline and 3 months. Baseline-only variables may be included as covariates when clinically appropriate, while follow-up-only variables, including adverse-effect items, will be summarized descriptively or analyzed as contextual outcomes. Within the XR-supported pathway, first-day to last-day changes across the 4-week intervention period will be analyzed as proximal pre/post outcomes, primarily using paired comparisons or mixed-effects models depending on the final data structure. Formal stratified analyses by tumor type, treatment characteristics, sex, or age group are not prespecified as confirmatory analyses because the expected subgroup sizes, particularly within the XR-supported pathway, may be small. Where the final sample size and data distribution permit, selected subgroup or interaction analyses may be conducted on an exploratory, hypothesis-generating basis.

Initial XR-specific design analyses will examine avatar representation and tutorial format within the XR-supported pathway. Avatar representation, defined as default versus participant-specific photorealistic avatar use, will be analyzed as a participant-level observational factor because uptake is not randomized. Tutorial format, defined as trainer-led versus body-double second demonstration, will be analyzed using a Bayesian modeling framework, consistent with previous virtual-body-double research using Bayesian modeling to estimate condition-specific responses and uncertainty (Gorisse et al., 2021). Because tutorial format is assigned at the routine-day level and linked to repeated exercise-level outcomes within participants, models will account for within-participant clustering and the repeated day-level exposure structure using hierarchical random effects. Continuous outcomes will be modeled using Gaussian likelihoods with identity links unless distributional diagnostics indicate the need for transformation or robust alternatives; binary outcomes will use Bernoulli likelihoods with logit links; ordinal ratings will use cumulative-logit models when the ordered response structure is retained; and count outcomes will use Poisson or negative-binomial likelihoods according to dispersion. Weakly informative priors will be used for regression coefficients and variance parameters, and model adequacy will be evaluated using convergence diagnostics, posterior predictive checks, and sensitivity analyses when required. Posterior summaries will include estimated effects, credible intervals, and posterior probabilities of direction.

Selected exercise intensity may be examined as an exercise-level behavioral outcome because it is chosen immediately after tutorial exposure. Bayesian models will assess whether selected intensity, completion, embodiment-related ratings, or movement-derived indicators differ between trainer-led and body-double demonstrations, and whether these patterns vary by avatar representation. Because selected intensity is ordinal, it will be modeled using an ordinal Bayesian approach if the distribution of responses supports this level of modeling.

Telemetry-derived variables, movement-derived variables, post-session ratings, tolerability indicators, and adherence or completion variables will first be summarized descriptively. Additional association analyses may examine whether these variables show coherent relationships with adherence, session completion, embodiment-related experience, tolerability, or patient-reported outcomes.

Missing data will be handled under a missing-at-random assumption. Linear mixed-effects models can accommodate incomplete repeated measures, and no formal imputation is planned for the main comparative analysis; complete-case sensitivity analyses may be performed. For Bayesian XR-specific design analyses, missingness, convergence, prior specification, and posterior predictive adequacy will be assessed according to the final model. Two-sided frequentist tests for primary and additional comparative analyses will use α = 0.05. XR-specific design and association analyses will be reported mainly through effect estimates, uncertainty intervals, posterior summaries, and descriptive patterns rather than binary significance decisions.

Data management and linkage

Data management combines paper-based assessments, web-based study data entry, and device-local digital records. The study-facing forms and guides described in Supplementary Data Sheets S1–S10 provide the human-facing layer of data capture and intervention delivery, whereas the local digital files described below provide the application-facing layer. Questionnaire data covering epidemiological, clinical, psychosocial, physical-activity, symptom, and wellbeing variables are collected through the study assessment forms and subsequently entered into a dedicated study-specific web platform by auxiliary study personnel. These records are linked to participants through the local study workflow.

In the XR-supported pathway, the application additionally generates two types of digital files. The first is a structured session log stored in JSON format under profile-based and session-based identifiers rather than personal names; it includes session timing, selected routine and day, avatar representation, exercise sequence, tutorial events, round-level intensity choices, rest periods, skip events, scene transitions, and exercise- or session-level scores. The second is a binary pose file containing the time-resolved positions and rotations of the retargeted avatar joints during use. A separate local profile file stores persistent usage variables such as app launches, avatar-selection history, routine-selection history, tutorial order group, and routine completion status.

The linkage strategy is therefore hierarchical. Participant-level study identifiers are managed through the local study workflow and the study web platform; profile-level identifiers organize the local headset data; and session identifiers distinguish repeated uses within the same profile. This structure allows questionnaire data, intervention exposure, event-based behavior, and movement-derived data to be linked for analysis while limiting the use of direct identifying information within the application itself.

Feasibility, data-quality outcomes, and progression criteria

The study evaluates not only patient-level outcomes but also whether NEOBIC-XR functions as a deployable and analyzable intervention platform. Feasibility and method-validation outcomes therefore concern onboarding completion, successful home initiation, prescribed-session adherence, repeated home use, tolerability, local logging, pose recording, technical-support requirements, and linkage between participant-level assessments and session-level digital records across centers.

Implementation indicators include completion of onboarding, successful home deployment, initiation of at least one home session, completion of prescribed sessions, interrupted or incomplete sessions, need for technical support or device replacement, early discontinuation, and uptake and successful deployment of the optional photorealistic-avatar workflow. Prescribed-session adherence is defined as the number of completed prescribed sessions divided by 12; partial sessions, replay sessions, and voluntary repetitions will be reported separately so that additional use is not conflated with completion of the prescribed dose.

Data-quality indicators include successful creation of local profiles, generation of structured session logs, availability of pose data when expected, completeness of key telemetry fields needed to reconstruct session progression, and successful linkage between questionnaire records, profile identifiers, and session identifiers. Telemetry completeness and pose-data completeness will be reported separately, because a session can be clinically completed even if one digital stream is incomplete or unavailable. Replay behavior, routine-selection patterns, tutorial skipping, round-level intensity choices, and session-duration distributions will be treated as contextual use descriptors.

Feasibility for a later confirmatory implementation study will be interpreted using the prespecified traffic-light framework shown in Table 2. These criteria are not intended to provide a single automatic pass/fail decision; rather, they identify domains that appear ready for progression, domains requiring modifiable remediation before progression, and domains indicating major feasibility concerns. A serious intervention-related adverse event will trigger independent clinical review and a progression decision rather than an automatic conclusion based only on numerical thresholds.

The anticipated results primarily concern feasibility, traceability, and analytic usability. The study is expected to show whether a multicenter, home-based XR exercise pathway for people with cancer can be delivered consistently enough to support translational refinement and later confirmatory evaluation.

At the methodological level, the study should indicate whether NEOBIC-XR functions as an analyzable intervention system. Key outputs include successful onboarding and home-use initiation, prescribed-session completion, interrupted or incomplete sessions, replay behavior, tolerability signals such as dizziness or nausea, need for technical support, availability of session logs and pose files, and successful linkage between questionnaire data and device-local records. Expected pitfalls include headset discomfort, cybersickness, tracking or Guardian-definition problems, incomplete post-session forms, delayed avatar deployment, and missing telemetry or pose data. The workflow addresses these risks through supervised onboarding, criterion-based home-use authorization, center-managed device preparation, participant support pathways, optional repeat training, and data-quality checks before interpretation of XR-specific analyses.

Within the XR-supported pathway, the study is also expected to generate initial evidence on two design dimensions: virtual body-double observation during the second tutorial demonstration and optional photorealistic avatar representation. Because intensity is selected immediately after tutorial exposure, selected intensity can be examined as a behavioral indicator alongside completion, embodiment-related experience, tolerability, and movement-derived variables. These analyses may identify preliminary patterns for later studies, but avatar-related contrasts will remain observational and tutorial-format analyses will be interpreted as initial within-XR evidence.

At the broader patient-outcome level, the study is expected to provide preliminary adjusted estimates of baseline-to-3-month change in self-reported physical activity and repeated patient-reported outcomes across the two pathways. First-day/last-day measures collected only in the XR-supported pathway will be interpreted separately as proximal within-pathway signals. Overall, the anticipated value of the study lies in determining whether the platform, deployment model, and linked measurement architecture are feasible, coherent, and informative enough to support future XR-based oncology rehabilitation research.

This Methods paper describes NEOBIC-XR as a home-based XR exercise platform for oncology supportive care and as a framework for evaluating selected design dimensions that are difficult to study in conventional rehabilitation delivery. Its main contribution is translational: it moves from the general recommendation that patients with cancer should remain physically active toward a structured, home-based, digitally supported pathway that can be deployed and measured under real-world conditions.

The platform treats XR not simply as a display for exercise prompts, but as a composite rehabilitation environment. Exercise is delivered through a tracked self-avatar, structured pre-exercise tutorials, adaptive intensity selection, mixed-reality support for selected exercises, and a center-managed home-deployment workflow. This combination is relevant in oncology, where symptom burden, physical deconditioning, reduced confidence in movement, and limited access to supervision often coexist. By combining a pathway comparison with nested analyses inside the XR-supported pathway, the study is designed to generate both implementation-level and design-level information.

A methodological strength is the multicenter, home-based deployment model. The workflow includes common procedures for onboarding, device provisioning, local profile handling, software updating, and optional deployment of participant-specific avatars across three hospitals. This center-managed and privacy-preserving workflow addresses a frequent barrier in translational XR research: not only developing an application, but distributing, updating, monitoring, and linking its data in a way that is acceptable for patients and clinical institutions.

A second strength is the layered measurement architecture. All participants contribute common baseline and 3-month patient-reported outcomes, whereas the XR-supported pathway additionally contributes first-day/last-day measures, post-session ratings, structured telemetry, and movement-derived pose data. This allows the intervention to be characterized in terms of delivery feasibility, adherence, tolerability, embodiment-related experience, proximal change during the XR-supported phase, and objective patterns of use and movement execution.

The study also has limitations. First, the main comparison is structured but non-randomized. Entry into the XR-supported pathway depends on participant acceptance and operational capacity, and uptake of the photorealistic avatar is participant-driven. Between-pathway and avatar-related contrasts may therefore reflect baseline differences in motivation, technology acceptance, symptom burden, or other factors associated with willingness to engage in XR-supported home exercise. Although the additional sensitivity analysis will assess robustness to measured baseline differences, residual confounding from unmeasured factors, including motivation or technology acceptance not assessed comparably before pathway allocation, cannot be excluded. These analyses should be interpreted as pragmatic and adjusted, not confirmatory.

Second, the intervention is multidimensional. Guided tutorials, self-avatar representation, optional photorealistic personalization, mixed-reality elements, flexible home use, and repeated digital feedback are delivered together. This is appropriate for a translational methods study, but it limits causal attribution to individual components. The nested design partially addresses this issue for tutorial format, which is counterbalanced, but not for photorealistic-avatar uptake, which remains observational.

Third, measurement density differs between pathways. The common comparative assessment is based on baseline and 3-month questionnaires, while session-level reports, telemetry, and pose data exist only in the XR-supported pathway. The study is therefore better suited to characterize the internal functioning and feasibility of the XR-supported pathway than to provide equally granular process-level comparisons between the two rehabilitation-support strategies.

Despite these constraints, the study may provide a useful foundation for future XR research in oncology rehabilitation. If NEOBIC-XR can be delivered across repeated home sessions, linked to clinic-based assessments, and managed across multiple hospitals with acceptable tolerability and data completeness, this would support more scalable XR-based supportive care strategies. In parallel, initial signals related to avatar representation, tutorial format, embodiment, tolerability, adherence, and movement execution may help refine future confirmatory studies with stronger allocation procedures and more focused research questions.

Overall, the present study provides an implementation and methodological foundation. It may help shift XR in oncology from a novelty-driven digital tool toward a structured rehabilitation-support strategy that can be evaluated, refined, and potentially integrated into real care pathways.