Longitudinal Assessment of Balance Using Virtual Reality in Patients Receiving Potentially Neurotoxic Adjuvant Chemotherapy

Christina Teng Concord Repatriation General Hospital Elodie Chiarovano The University of Sydney David Gregory Tighe University of New South Wales Kim Tam Bui Concord Repatriation General Hospital Venkatesha Venkatesha Sydney Local Health District Prunella L Blinman Concord Repatriation General Hospital Hamish Gavin Macdougall The University of Sydney Janette L Vardy (  janette.vardy@sydney.edu.au ) Faculty of Medicine and Health, University of Sydney; Centre for Medical Psychology and EvidenceBased Decision-Making, The University of Sydney; Concord Cancer Centre, Concord Repatriation General Hospital. https://orcid.org/0000-0002-5739-5790


Introduction
Adjuvant chemotherapy improves survival in patients with breast and colorectal cancers. Chemotherapyinduced peripheral neuropathy (CIPN) is a common side effect of agents used in the treatment of many common cancers. CIPN has a cumulative dose-dependent relationship with chemotherapy, limits the amount of adjuvant chemotherapy that can be safely delivered, and is a common reason for early cessation or dose reduction of treatment. CIPN can cause progressive sensory disturbances or pain in the Our study objectives were to: 1) assess and quantify the effect of neurotoxic chemotherapy on balance using VR; 2) explore whether lower limb neuropathy assessed by validated clinical and patient-reported tools was associated with VR assessment of balance; and 3) to determine the safety and tolerability of VR assessment in the oncology clinical setting.

Study design
This prospective, longitudinal observational study assessed balance using VR and CIPN using multimodality neuropathy assessment tools in patients receiving platinum-or taxane-containing adjuvant chemotherapy for bowel or breast cancer.

Setting
Participants were recruited from two Australian tertiary oncology centres in New South Wales, Australia from November 2018 to July 2020. This study was approved by the Human Research Ethics Committee of Sydney Local Health District (Concord Repatriation General Hospital, HREC/18/CRGH/197) and conducted according to the Declaration of Helsinki. All patients provided written informed consent before study participation.

Participants
Eligible participants were chemotherapy naive, over 18 years of age and planned to receive adjuvant therapy containing docetaxel or paclitaxel for breast cancer, or oxaliplatin for colorectal cancer. Participants required su cient English to complete study assessments. Speci c exclusions were intolerance of VR, inability to stand unassisted in a stationary position, and pre-existing balance disorder such as myelopathy, Meniere's disease or Parkinson's disease. Participants with diabetes but without clinical neuropathy were eligible.
Patient and treatment characteristics extracted from the medical le included: age, cancer diagnosis, height, weight, chemotherapy regimen, and reasons for chemotherapy modi cations (dose reductions, delays or early cessation).

Data sources/measurement
Assessments were performed at baseline, pre-speci ed intervals throughout chemotherapy, and 3-and 6months following completion of chemotherapy. Assessment intervals throughout chemotherapy were scheduled to coincide with clinical reviews, which were every 3-6 weeks depending on the length of the chemotherapy cycle. After March 2020 and in the context of the COVID-19 pandemic, some participants had telehealth instead of face-to-face appointments; hence, some assessments were unable to be completed.
Balance assessment using VR Postural sway was measured by a Wii Balance Board (WBB) connected by Bluetooth to an iPhone running the BalanceRite App. The participant was instructed to stand on the WBB or on the WBB plus a foam pad (WBB + foam, Airex AG, Sins, Switzerland, 41 cm × 50 cm × 6 cm thick) and to stay as still as possible for 20 seconds. Several visual conditions were presented: eyes open, eyes closed, and with an unpredictable visual perturbation VR at several amplitudes of movement. The visual perturbation consisted of a stereoscopic visual scene (Tuscany garden) projected into an Oculus Go VR headset. The visual perturbation was graded into levels from 0 (no perturbation) to 5 (highest amplitude of perturbation). Using eyes open, eyes closed, levels 0 to 5, with and without the foam pad, we effectively created 16 levels of di culty in testing conditions [14]. The equipment is demonstrated in the supplementary material S1.
Each participant was assessed for their maximum tolerated VR level of di culty (their 'balance threshold'). Stepping off the WBB, or requiring external support for balance constituted reaching the 'balance threshold'. At subsequent assessments, participants were initially tested at the balance threshold. If they were unable to maintain their balance at this level, the di culty was progressively reduced until they were able to complete the level, and this became their new balance threshold. This protocol aimed to limit participant exposure to the testing environment, to avoid improving on the balance task performance due to a repetition/training effect. With each participant acting as their own control, the change in balance threshold level from baseline was determined at subsequent timepoints. 3. Clinical neuropathy using The Total Neuropathy score (clinical) (TNSc), a composite test comprising clinical examination including muscle weakness, pinprick and vibration sensation, tendon re exes, and clinician-rated sensory, motor and autonomic symptoms. It is used as a clinical measure of CIPN severity with higher scores indicating greater neuropathy (range 0-28) [19].
4. Falls risk, using the validated 'Timed Up-and-Go' test. The score is the time taken for a participant to stand from a seated position, walk three metres to a marker, and return to their seat [20].
5. Patient-reported fall or near-fall (fall event) since the previous assessment.

Study size
The sample size of 35 participants was pragmatic based on anticipated uptake over a recruitment period of 18 months.

Statistical analysis
The primary endpoint was the difference in the measured balance threshold between the baseline and end-of-chemotherapy assessments. Descriptive statistics were used to compare scores for grade of neuropathy, FACT/GOG-Ntx13, TNSc, and 'Timed Up-and-Go' between baseline, end-of-chemotherapy and 3-and 6-month post treatment time-points.
The normality of the data was assessed using Normal Q-Q plot. Interval scale outcomes (FACT/GOG-Ntx13, TNSc, 'Timed Up-and-Go') were compared using repeated measures ANOVA. In case of nonnormality, Friedman's test was applied. The binary outcome (fall event yes/no) was compared using Cochran's Q test followed by McNamara's test with Bonferroni's correction for post-hoc comparisons.
Ordinal outcomes (CTCAE grade of neuropathy and VR balance threshold) were compared using Friedman's test followed by Wilcoxon's test with Bonferroni correction for post-hoc multiple comparisons.
To account for patients with missing data, multiple imputation was used as recommended for the analysis of outcomes in longitudinal studies [21,22]. Independent t-tests (Mann-Whitney U test) were used to compare VR balance threshold between participants who had a fall event, and those who did not. All statistical tests were performed at 0.05 level of signi cance. All analyses were performed using SPSS V26. Due to the limited sample size and the exploratory nature of the study, subgroup analyses were not performed.

Participants
Of 35 participants who consented, 34 completed the baseline assessment and 32 completed at least two assessments while on chemotherapy. Most participants were female, who received adjuvant taxane chemotherapy for breast cancer. The demographic details of participants are summarised in Table 1.
Treatment intensity was > 75% of planned in 32/34 participants. Dose modi cation or early cessation of chemotherapy occurred in 26/34 (76%) patients. CIPN was the most common reason reported, affecting 50% of patients regardless of cancer diagnosis. Two participants only achieved 50% of planned treatment; one due to fever and infection, and one chose to stop chemotherapy due to the COVID-19 pandemic.
Outcome data Participants completed a median of 4 assessments (range 2-5), dependent on their chemotherapy cycle length and duration of adjuvant therapy. Figure 1 outlines the ow of participants through the study.
The primary endpoint was evaluable for 28 participants. Reasons for missing data in 5 patients are: 1) 3 participants declined the VR assessment due to fatigue; 2) no assessor was available for 1 assessment; and, 3) one patient relocated treatment to a different hospital.
The reduced number of assessments at the 3-(n = 22) and 6-(n = 23) month post chemotherapy timepoints were predominantly due to reduced face-to-face visits in the context of the COVID-19 pandemic.
Primary endpoint: VR balance threshold A reduction in balance threshold was observed in 10/28 (36%) of participants assessed at the end of chemotherapy, indicating that they could not maintain their balance at the same level of challenge as prior to chemotherapy. Most patients reduced their balance threshold by one level, but one participant patient dropped by 12 levels. Four participants (15%) had a balance threshold reduce by greater than two levels.  Results of the repeated measures ANOVA for complete cases data indicated no signi cant difference in the 'Timed Up-and-Go' across the four assessments (n = 9, Wilk's lambda = 0.57, F (3, 6) = 1.5, p = .31).

Frequency of falls
There were no participants or researcher injuries relating to the VR assessments. A fall/near fall was reported by 12 of 34 patients during the study period. Seven patients had falls, three had a near fall, and two had both. Three patients reported multiple episodes. Of the 18 falls or near falls reported, 13 occurred while on chemotherapy, two were within three months of completing chemotherapy, and three were between 3 and six months of treatment completion.
Balance threshold measured following the reported fall/near fall was signi cantly lower compared to times when no fall was reported (mean 10.3 vs 14.2, p value = 0.0002). Similarly, scores for clinical and patient reported neuropathy showed signi cantly more impairment from CIPN in participants who had experienced falls/near falls, compared to when no fall was reported (Table 3).

Discussion
This prospective cohort study used VR as a novel assessment of balance in people receiving potentially neurotoxic adjuvant chemotherapy. Balance de cits were identi ed in 36% of the participants at the completion of chemotherapy, and often persisted at 6 months following chemotherapy. The balance assessment ndings were consistent with worsening CIPN identi ed by clinician-and patient-reported measures, which were also signi cantly and persistently higher following chemotherapy compared to baseline. Falls risk measured by 'Timed Up-and-Go' did not change signi cantly over time.
Participants who experienced a fall or near fall during the study (35%) were more likely to have balance de cits when assessed using VR assessment; they also had signi cantly higher CIPN scores assessed by clinician-grade, patient-reported outcome questionnaires and clinical parameters.
The ndings of CIPN at 6 months following chemotherapy in 72% of our cohort is consistent with the prevalence of 58% (95% CI: 42-73) reported by our recent meta-analysis pooling prevalence of CIPN reported from clinical trials of adjuvant oxaliplatin [23].
A recent review has highlighted the negative impact of CIPN on physical function, contributing to postural instability and falls risk [24]. Prior studies examined standing balance in patients treated with chemotherapy using different tools and different populations. Balance de cits were linked with CIPN severity but authors did not report on near or actual falls [25,26].
This is the rst study to explore the use of VR technology to assess balance in the oncology clinical setting, and compare the results with near or actual falls. VR technology has been utilised in oncology care with a systematic review indicating bene t in inpatient rehabilitation, and as a distraction therapy to alleviate pain and anxiety around procedures [16].

Strengths and limitations
The main strengths of this study are its prospective longitudinal design, and 'real world' cohort of participants with common cancer diagnoses requiring adjuvant chemotherapy. These allow for applicability of the ndings to the clinical context. It assessed balance in a novel way using readily accessible equipment, and is 'proof of concept' of how VR technology might be integrated into the clinic setting. Feedback from clinicians was not a prespeci ed outcome of this study, and would be required to determine the role of VR as an assessment tool in a clinic setting.
The limitations of this study are the modest sample size, and missing data in the follow up period due to the uptake of telehealth consultations in the context of the COVID-19 pandemic. This limits the ability to perform subgroup analyses, and the precision of our point estimates. There is the potential for bias given the missing data; if participants refused assessments due to feeling unwell, side-effects from either chemotherapy of VR use may be underreported. As this was an observational study, feedback from the CIPN measures and balance assessments performed as part of the study was not fed back to the treating clinicians in real time. A further interventional study validating the test and evaluating the utility of treatment modi cations on the basis of the VR assessment would be required prior to applying an individual's results to changes in clinical care. Our study population did not include patients with advanced disease or who received other cancer treatments with the potential to cause neurotoxicity such as cisplatin, bortezomib, or nab-paclitaxel. These factors need to be considered when applying our ndings.

Interpretation
This novel test is simple, safe, uses equipment that is easily accessible. Its utility in identifying CIPN was no better than the existing assessment strategies available. However, it allows for an objective assessment on the functional disability of CIPN on an individual's falls risk. It would assist with identifying functional limitations when patients are unwilling to disclose symptoms, either due to a lack of insight or a fear of ceasing chemotherapy early. These and other factors were identi ed in a qualitative study exploring reasons for under-reporting of CIPN [27]. Poor correlation between patient self-report of balance and objective balance testing has been described in a prior study using the same VR equipment and methods used in this study [15].
Rather than being limited to CIPN measurement, VR balance assessment is likely to measure a multitude of falls and balance risk factors which are not necessarily captured solely by CIPN tools. Peripheral nerve function plays a small but important part in maintaining balance. Other factors such as sensory input, cognitive function and physiological aspects also play an important role [28,29]. Each of these factors can be impacted by chemotherapy treatment in different ways and affect the overall falls risk.
Exercise has shown to be bene cial in the prevention and treatment of CIPN [30,31]. VR exercise programs offer the potential for a patient-led physical activity with data feedback to clinicians or exercise physiologists. VR may be a useful tool for the prevention or rehabilitation in populations at risk of falls, regardless of the aetiology.

Conclusion
Impaired balance threshold using VR was associated with actual or near falls in patients receiving neurotoxic chemotherapy. VR is safe and provides objective measurement of falls risk in the oncology clinic setting, although in our population was no more effective at identifying CIPN than existing measures. Further research into the use of VR as a prevention or rehabilitation strategy in patients at risk of falls is warranted.

Declarations
Funding: Funding was provided by the M.A.S.T. foundation to purchase VR equipment, and support a research assistant for data entry. Dr Janette Vardy is supported by a National Health Medical Research Council Investigator Grant (APP1176221).
Con icts of interest: None to declare.
Availability of data and material: The authors have full control of primary data. Any data sharing will be at the discretion of the corresponding author following a written request.
Code availability: Not applicable.
Authors contributions: CT, PB, JV developed the study concept and protocol. EC and HG developed the VR assessment strategy, and instructed the research team on the use of the equipment.
CT, EC, DT and TB collected data and performed participant assessments. CT and V performed data analysis.
Ethics approval: This study was approved by the Human Research Ethics Committee of Sydney Local Health District (Concord Repatriation General Hospital), (HREC/18/CRGH/197) and performed in accordance with the ethical standards of the Declaration of Helsinki.
Consent to participate: All participants provided written informed consent before study participation.
Consent for publication: All participants have provided written consent for their de-identi ed data to be publishes, and all authors have provided consent to publish the manuscript.   Neurotoxicity questionnaire-13 items. a. higher score indicates worse symptoms, b. higher score indicates worse neuropathy. Figure 1 Participant ow diagram.

Figure 2
Outcome measures of participants. A) Highest balance level reached using virtual reality equipment (higher score indicates better balance); B) Functional Assessment of Cancer Therapy/Gynaecologic Oncology Group (FACT/GOG) neurotoxicity questionnaire (higher score indicates more symptoms); C) Timed up-and-go (higher score indicates impairment); D) Total Neuropathy Score (clinical) (higher score indicates impairment)

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.