Sarcopenia is characterised by the age-related decline in muscle mass plus loss of muscle strength and/or reduced physical performance and can result in increased vulnerability to adverse health-related outcomes.1 2 Secondary sarcopenia refers to muscle health decline related to acquired impairments resulting from disease, trauma, treatment complications or degeneration.3 The timely identification and risk stratification of sarcopenia in at-risk populations with relevant comorbidities will potentially allow early interventions to avert the downstream deleterious consequences of frailty such as functional decline, mobility deterioration, falls, fragility fractures, hospitalisation and mortality.4

Comorbidities that confer increased risk of sarcopenia include both acute and catastrophic diseases such as stroke, traumatic brain injury (TBI) and breast cancer, as well as chronic diseases such as knee osteoarthritis (OA). Previous studies have alluded to the presence of sarcopenia at disease onset as a negative predictor of later functional outcomes. Stroke-related sarcopenia begins early after stroke onset, and its presence on acute hospital admission has been shown to increase the risk of poor functional outcomes at 6 months.5 In TBI, the prevalence of reduced lean skeletal muscle mass on acute admission, as determined by cross-sectional CT imaging at the third lumbar vertebrae, was significantly associated with worse global outcomes using the Glasgow Outcome Scale.6 Despite curative treatments and disease remission, breast cancer survivors commonly experience low physical activity levels and are at increased risk of sarcopenia.7–10 Patients with knee OA are at risk of inactivity, with some studies suggesting a relationship with sarcopenia, although it is uncertain which condition preceded the other as these studies were largely observational and cross-sectional/case-control in nature.11 12

In particular, osteosarcopenia, defined by the combination of low bone density (osteopenia/ osteoporosis) and muscle mass, strength and/or functional capacity (sarcopenia),13–15 has been described in a subset of patients. In a meta-analysis of sarcopenia and associations with falls and fractures, it was found that the prevalence of sarcopenia in patients with low-energy fracture (n=9) was 46%, with the mean bone mineral density and T-score found to be significantly lower in sarcopenic patients.16 However, the study settings were very diverse due to heterogeneous populations, comprising healthy volunteers, volunteers with increased risk of fall, participants referred to outpatient clinics treating osteoporosis and preventing fall and patients with verified low-energy fractures, with age cut-offs ranging from >50 to >70 in different studies. While many studies have been performed in western cohorts, limited studies have been performed in Asian patients. In a study of healthy community-dwelling volunteers in Japan with a mean age of 71.4 years, the prevalence of osteosarcopenia was estimated to be 8%, osteoporosis alone to be 14% and sarcopenia to be 13%, with findings of significantly lower BMI and back muscle strength in this group.17 In a Korean study recruiting high-risk patients with postoperative hip fractures with a mean age of 83.0 years, there was a prevalence of osteosarcopenia of 46%, and those with preoperative osteosarcopenia were found to have poorer functional outcomes and higher fracture incidence within 1 year post surgery.18 Another study in Singapore found a prevalence of 11.7% of osteosarcopenia in a cohort of independent, community-dwelling elderly individuals (mean age of 71.4 years in the osteosarcopenic group compared with a mean age of 65.7 in the normal group), with nutrition mediating the association between osteosarcopenia and frailty.19 In addition, patients with osteosarcopenia were found to have a mean fried frailty total score of 1.7, compared with 0.38 in the normal group. However, it is still unclear how osteosarcopenia acts singly or in concert with other clinical factors to influence the expression and trajectory of the frailty continuum, especially in the high-risk cohorts in our study.20

Studying the evolution of muscle changes is important as it can affect functional prognosis and rehabilitative outcomes, such as gait, limb function and swallowing ability. Traditional modalities have been used, such as dual-exergy X-ray absorptiometry (DEXA), CT, MRI, although these have largely been confined to academic settings.

More recently, increasing recognition of the clinical utility of muscle ultrasound (MUS) in these populations has emerged, given the relative affordability, portability, access and lack of ionising radiation as compared with traditional modalities. Systematic reviews have shown that ultrasound is a reliable and valid tool for muscle assessment in older adults,21 for sarcopenia assessment22 and to predict clinical outcomes including functional status at discharge, nutritional status and length of stay.23 MUS allows the study of muscle changes in clinical settings with minimal disruption to patients undergoing inpatient rehabilitation or receiving outpatient care. Small studies of MUS in stroke patients have shown changes in the upper and lower limb muscles which start early after stroke onset,24 25 and with an association found between limb muscle architectural parameters and ambulatory outcomes.26 However, early MUS and longitudinal outcomes have not been well documented in patients with TBI, breast cancer and knee OA.27 Additionally, there is also a lack of data on MUS changes and associations in Asian patients.

This study aims to investigate the prevalence and longitudinal evolution of osteosarcopenia in patients with stroke, TBI, breast cancer and knee OA. Additionally, we aim to study the associations between the respective acute disease characteristics, biomarkers (sarcopenia, muscle, bone, MUS, fluid biomarkers and physical activity) and their impact on global functional outcomes and health-related quality of life.

This is a multicentre, prospective, longitudinal observational cohort study that is designed to investigate multidimensional biomarkers of osteosarcopenia and physical activity in patients with stroke, TBI, breast cancer and knee OA.

Patients with stroke, TBI, breast cancer and knee OA will be recruited from the Tan Tock Seng Hospital Rehabilitation Centre and Tan Tock Seng Hospital. Tan Tock Seng Rehabilitation Centre is a tertiary level rehabilitation unit in Singapore (a city-state), which receives referrals from National Healthcare Group hospitals. Tan Tock Seng Hospital is an acute tertiary-level hospital in Singapore. The study is planned to start in January 2024 and end in January 2027.

Study inclusion criteria are as follows: (i) age ≥40 years, (ii) Asian ethnicity based on hospital admission identification records, (iii) first confirmed diagnosis of either stroke, TBI, breast cancer or knee OA, (iv) living in community and (v) ability to understand one step simple commands.

In addition, for inpatients, (i) admission was within 12 weeks of stroke/TBI onset and (ii) rehabilitation admission was within 2 weeks. The rationale for these criteria is to allow clinical measures to be obtained during the acute clinical phase and based on clinical populations validated by the Asian Working Group for Sarcopenia (AWGS) 2019 consensus.28

For outpatients, additional inclusion criteria are as follows: (i)>6 months from initial diagnosis of first stroke, TBI, breast cancer or knee OA, and (ii) at least standby assistance, modified independent or independent in ambulation with/without walking aids. Hence, outpatients with recovered conditions with no apparent frailty, early or prefrailty will be recruited for this study, that is, clinical frailty scale CFS 1–5.29 Patients were recruited >6 months after their initial diagnosis so that clinical measures can be obtained during their subacute/chronic phase of their disease condition.

Patients will be enrolled after written informed consent is obtained and assigned an enrolment number. Inpatient and outpatient clinic lists will be screened by research assistants for potential eligible patients. Inpatients will be screened on admission to the rehabilitation centre. The research assistants will then obtain consent from the primary physician before approaching the patient or legal representative for informed consent. Clinical data will be obtained from hospital records, and additional research information will be collected on hardcopy. Recruitment period is anticipated to end on 31 Oct 2026. The timeline of recruitment and data collection is stated in online supplemental appendix 1.

The primary outcome is osteosarcopenia. Osteopenia and osteoporosis are classified by the WHO criteria based on the BMD T-score of the lumbar spine and/or femoral neck and/or total hip equivalent of −1 to −2.5 SD and lower than −2.5 SD, respectively.30 Sarcopenia was diagnosed with the AWGS 2019 consensus criteria, defined as low relative appendicular skeletal muscle mass in the presence of either low handgrip strength or slow gait speed.28 Osteosarcopenia was defined as co-existent osteopenia/osteoporosis and sarcopenia.

Secondary outcomes are physical activity and physical performance variables. These outcomes will be captured through questionnaires, physical performance assessments, and muscle and bone imaging modalities. Blood samples will be biobanked for future analysis. These are described in detail below. Multiple site initiation meetings have been arranged to develop and standardise the standard operating procedures among the study team.

Clinical and demographic characteristics will be collected from participants. For stroke patients, these will include National Institutes of Health Stroke Scale scores,31 location of stroke, treatment and functional scales. For TBI patients, these will include Glasgow Coma Scale on arrival, treatment and functional scales. For knee OA patients, these will include the Kellgren– Lawrence radiographic grading32 and functional scales. For breast cancer patients, these will include the cancer stage, type of surgery and adjunctive treatment. This will allow stratification of disease severity and treatment during data analysis.

Clinical data about participants' demographic, socio-economic, housing and educational status will be obtained.

Frailty states will also be screened using the FRAIL sore, which is a five-question assessment of fatigue, resistance, aerobic capacity, illness and loss of weight, which classifies patients into the category of non-frail, prefrail and frail.33 Additionally, they will also be screened with the clinical frailty scale.34 This is a clinical judgement-based frailty tool, which evaluates specific domains including comorbidity, function and cognition to generate a frailty score ranging from 1 (very fit) to 9 (terminally ill).

Multimorbidity will be measured using the Charlson comorbidity index.35 Malnutrition will be assessed with the simplified nutritional appetite questionnaire, which is a validated four-item self-reported nutritional screening tool with total score ranging from 4 to 20, with a cut-off of ≤15 being a marker of malnutrition risk and involuntary weight loss.36 Self-reported activity level is assessed with IPAQ-E. This is a self-reported recall questionnaire that assesses daily physical activity of older adults, based on the past week, and further subdivided into sedentary, light, moderate and vigorous activities.37

Quality of life is assessed with the EuroQoL 5D questionnaire of five levels (EQ5D-5L).38 39 It has been validated in many countries including Singapore.40 The instrument consists of five questions, each assessing problems in one of the five dimensions: mobility, self-care, usual activities, pain/discomfort and anxiety/depression. Each dimension is assigned a level using a five-point scale: no problems, slight problems, moderate problems, severe problems or inability. EQ-5D-5L also includes a 100-point visual analogue scale for respondents to assess their general health.

Anthropometric measurements for each participant will be measured by trained personnel. Blood pressure, body weight, height, waist and calf circumference will be obtained. Grip strength will be measured using a hand dynamometer (Jamar Plus Digital hand dynamometer, Jamar North Coast Medical Inc., Morgan Hill, CA, USA). Using two trials of grip strength obtained for each hand, we will use the maximum handgrip strength reading that has been shown to have better predictive validity for poor physical performance.41

Physical performance will be measured using the short physical performance battery (SPPB), which comprises balance, gait speed and five times-chair stand test, with a total score cut-off of <10, indicating poor physical performance.42

MUS will be performed for four separate muscles using standardised positioning and measurement techniques for the following bilaterally: masseter, biceps brachii, rectus femoris and medial gastrocnemius. To assess muscle mass and quality, we will measure muscle thickness and echo intensity respectively. This will be done using B-mode imaging. A single MUS will be used for all participants and measurements (Resona 7, Mindray, Shenzhen, China) and all measurements will be taken by two experienced clinician sonographers, JMK and MRJT.

Ultrasonographic evaluations of the following muscles will be performed with patients in a supine position in the following manner.43 The masseter is measured with the probe perpendicular to the anterior margin of the masseter and external surface of the mandibular ramus, between 2 and 2.5 cm above the lower mandibular margin in a supine position with molars of both arches touching without pressure.44 The biceps brachii is measured 66% of the distance from medial acromion to elbow crease.45 The rectus femoris is measured midway between the anterior superior iliac spine and femoral lateral epicondyle.46–49 The medial gastrocnemius is measured 30% proximal between popliteal fossa and posterior calcaneus, with the ankle in relaxed position.50

Muscle thickness (mm) is derived from the measurement of the longest distance between the superficial and deep aponeurosis where applicable. Muscle echo intensity is determined through selection of a region of interest and calculating the mean echo intensity, expressed as a number between 0 and 255 arbitrary units (AU), where 0 represents black and 255 represents white.51 Muscle thickness and echo intensity were assessed using ImageJ software (National Institutes of Health, Bethesda, USA).52

Each patient will undergo BMD and dual-energy X-ray absorptiometry (DXA; Hologic QDR 4500, Hologic Inc., Waltham, MA, USA or equivalent) once at baseline during the study period. All scans will be performed using standard procedures as described in the User’s Manual. BCA via DXA will be used to determine lean muscle mass (g) and will be done once at baseline during the study period. Measurements will be taken with all subjects in a lying position and, if necessary, strapping for the affected upper/lower limb in stroke/TBI patients will be done to aid positioning for those with limb asymmetry due to spasticity, muscle imbalance or joint contractures. The total bone mineral density (g/cm²) will be determined for lumbar spine, hip and femur bone.

In-clinic BCA will also be assessed by a bioimpedance analysis (BIA) device (InBody 770; InBody Corp, Seoul, South Korea) at each outpatient study visit at 6 months, 1, 2, and 3 years) for participants who are able to stand independently. Appendicular segmental lean mass index (ASMI) (height-adjusted ASM (ASM/height²)) will be investigated in this study. For neurologically impaired patients with hemiplegia or post-stroke spasticity, axillary spacers or hand bandages will be used to ensure optimal positioning and grip for the BCA hand sensors as needed.

Baseline blood specimens will be tested for (i) metabolic markers comprising LDL, HDL, TG, total cholesterol, HbA1c, fasting glucose and 25-OH vitamin D and (ii) inflammatory markers (C-reactive protein (CRP), albumin, TNF-alpha and monocyte chemoattractant protein-1 [MCP-1]). Metabolic markers will be processed immediately and the results will be available in the hospital records, which will be retrieved by the study team. Inflammatory markers will be stored and processed in batches by the laboratory. (Online supplemental appendix 1)

Participants will be issued a wristband accelerometer (Fitbit Inspire 3, Google, CA, USA) for their own use over 2 weeks, and their daily step counts over that period will be recorded.

Markerless artificial intelligence-driven motion analytics via multiple cameras and single camera on web browser will be developed to measure five times-chair stand test, gait speed and components of SPPB. These will be validated against traditional gait measures in healthy participants and participants with neuropathological gait. The markless motion analytics used have been validated previously.53 54

All questionnaires, data collection and procedures will be supervised and administered by trained research staff (research assistants or physiotherapists). Multiple site initiation meetings have been arranged to develop and standardise the standard operating procedures among the study team. DXA and BMD will be performed by specialist radiologists at Tan Tock Seng Hospital.

The timeframe of measurements is displayed in table 1 for inpatients and table 2 for outpatients. For outpatient assessments, patients will be called back to the hospital.

The sample size calculation used in this study was an inequality test for two proportions to compute power and sample size for estimated disease prevalence (ranges) and hypothesised effect size (ie, OR).55 We assumed that the prevalence of osteosarcopenia (ie, primary outcome) would vary between 20% and 30% in different rehabilitation populations such as stroke and TBI, and recovering cancer breast cohorts will vary from 30%.56 57 We also hypothesised that the estimated effect size (OR) will be 2.0. The test statistic used is the two-sided Z test with pooled variance. A significance level of 5% and a power of 80% were applied in this calculation. A total of 280 subjects will be required. Further assuming an 8% attrition rate, a total of 302 subjects will be required (table 3). The above hypothesised figures were based on local data with a conservative approach using minimal detectable effect size, which will be sufficient for a higher anticipated detectable effect size (OR) of >2.0. PASS software V. 16 (NCSS LLC, Kaysville, UT) was used for sample size calculation.

Continuous data will be described as mean and SD for normally distributed data and median values with minimum (min), maximum (max) for non-normally distributed continuous data. Log transformation will be performed for the continuous data with skewed distributions. Categorical variables will be presented as numbers and percentages with 95% confidence intervals; χ² test and Fisher’s exact test will be used to evaluate differences. X2 test for trend will be applied to test for trends.

The prevalence rates of osteosarcopenia (ie, primary outcome) will be presented as crude rates, disease domain-specific rates and adjusted rates with 95% confidence interval. Bivariate (univariate) analysis will be conducted to identify putative predictor variables associated with the outcome measure (osteosarcopenia). Multivariate regression analysis by backward stepwise logistic regression procedure will be performed to adjust for putative confounding variables, including clinically important variables in the model.

P ≤ 0.1 with ORs that exclude 1 will be used as the cut-off for statistical significance for variable selection for multivariate modelling in order not to miss any potentially important predictors. Statistical significance remains the conventionally defined p≤0.05 and ORs that exclude 1 in the bivariate and multivariate models. To choose among competing models, the preferred logistic regression model will be selected, based on the log likelihood ratio. To evaluate model fitness, we performed the Hosmer–Lemeshow goodness-of-fit test.

In view of the uncertainty attached to the anticipated effect sizes of primary outcome measures, an independent data monitoring committee will be established to review the interim results of the study together with other evidence which may become available. When approximately 30%–50% of enrolled participants have completed the study, the committee will conduct an interim analysis for initial assessment of outcome data, safety data and sample size. The interim data will be used to review the feasibility of the study and to justify the sample size. Based on the findings, variables with minimal correlation may be excluded from further data collection. The data for such a review will remain confidential to the committee members and the study statistician, who will forward their recommendations to the Steering Group and take the necessary action.

The data sets obtained from this study will be available from the corresponding author at the end of the study on reasonable request.

With intentionally wide inclusion criteria, deep phenotyping, socio-economic data and longitudinal follow-up, we believe that our results will have good internal and external validity for the studied index conditions and provide valuable insights into relationships between acute disease characteristics, baseline muscle health, bone, muscle and blood biomarkers and subsequent evolution of osteosarcopenia.

Through physical activity tracking and gait metrics, we hope to study the relationship between step counts and sedentary time on sarcopenia severity and its subsequent progression. We will be able to study the relationship between clinical and imaging biomarkers of muscle in populations with various impairments and disabilities. Findings of affected muscle architecture have been reported even in early stroke patients, and we aim to document any changes in muscle morphology during the early phase of hospitalisation in our study population.24 There appears to be consistent evidence that lower limb MUS parameters are associated with ambulatory outcomes in neurological and non-neurological patients,49 58 with conflicting findings for MUS in the non-weight bearing upper limb.25 59 Masseter MUS may also play a role as a nutritional biomarker.60 Hence, investigating these key ultrasound regions in our study population will provide insights into the longitudinal muscle changes and health correlates in our study population.

The findings from our study will enable the development of risk stratification indices, which can be developed to determine the risk of developing poor intermediate and longer-term functional outcomes; these in turn will guide the development of risk-stratified individualised interventions to shift the frailty curve to the left.

This study will provide novel longitudinal evidence of muscular and metabolic adaptation to neurological impairments, and the influence of physical activity or the lack thereof, which has not been well studied. While physical activity has been demonstrated to reduce secondary complications of disability due to stroke and TBI, sedentary behaviours are often the norm. Hence, results from this study may inform the physical performance and activity level of our cohort with a potential to drive future effective measures to address such health deficits to reduce cardiometabolic risk factors and osteosarcopenia and the associated burden to public health resources.61 62

We highlight some methodological limitations. There may be missing data points from patients who are unable to perform physical tasks due to severe stroke-related physical, cognitive or language-related impairments (eg, post-stroke aphasia and TBI-related attentional impairments). However, these patients will continue to receive other measurements such as MUS, BMD and blood biomarkers which can be done in the supine position with minimal active participation. We will also accommodate physical impairments where possible, for example, adaptations for hemiplegic postures to minimise overlapping joint positions or improve foot/finger sensor contact for BIA measurements to mitigate the effects on the validity of readings. Due to the duration of assessments, study duration and follow-up periods, this may negatively affect recruitment rates, induce participant fatigue and give rise to drop-out rates. This is mitigated by appropriate rest breaks during assessment sessions and participant reimbursement for their time, effort and partial defrayment of transportation costs. Due to resource constraints, measures related to cognitive frailty domains were not studied. In addition, as our study excludes patients who have impairments affecting understanding of questionnaires and tasks, our study population may exclude patients with cognitive frailty.

Not applicable.