This study is designed as a prospective, experimental, multicenter, 3-cohort investigation. The research will be performed at Albacete University Hospital Complex, Spain (with the coordinating Clinical Research Ethics Committee; hospital 1); Getafe University Hospital, Spain (hospital 2); and primary care units of Donostialdea, Osakidetza, Spain (primary care units). The primary health care units involved in the study will be located in 2 districts/regions of Gipuzkoa: the region of Donostialdea and the region of Tolosaldea. The design of this study protocol, characterized as exploratory, adheres to the SPIRIT (Standard Protocol Items of the Recommendations for Interventional Trials; Multimedia Appendix 1) 2013 guidelines, as depicted in Figures 1 and 2.

The allocation sequence in this multicentric study is institution-based. Patients recruited at the Department of Geriatrics of the Complejo Hospitalario Universitario de Albacete will be automatically allocated to the supervised exercise program by default. By contrast, patients recruited at the Geriatrics Department of the University Hospital of Getafe will not be assigned to any specific interventions. Similar sociodemographic characteristics are anticipated for both patient populations. At each institution, patients will be recruited from the falls unit, outpatient clinics, and day hospital. Within each acquisition site, patient selection will be conducted based on the clinical agenda until the recruitment goals are met.

The allocation sequence is sequential, aligned with the clinical agenda, and overseen by the clinical principal investigator at each site. Any deviations from the recruitment plan outlined in the clinical agenda must be thoroughly justified and meticulously documented.

In the hospital exercise cohort, a 16-week supervised multicomponent physical exercise program will be implemented. The individualized exercise intervention, tailored to each individual’s functional capacity, comorbidities, and previous experience, will consist of 2 supervised sessions per week, each lasting 45 minutes. These sessions will be conducted in small groups, with 4-6 older adults per group. The sessions are divided into warm-up, main part (strength, power, and coordination exercises), and cool down (flexibility and stretching). The main part of the exercise program comprises 8 exercises targeting the major muscle groups. The progression starts with low-intensity and high-volume sessions, gradually advancing to higher intensities and lower volumes. Depending on the stage, participants will complete 2-4 sets, consisting of 4-15 repetitions, with a 30-second rest between sets and exercises. For each exercise, loads and intensities will be adjusted to achieve a total of 30 repetitions. All exercise sessions are to be conducted in the gym of the Geriatrics Department of the hospital. Participants will be encouraged to incorporate additional 90-minute walks per week into their routine. It is important to note that the exercise program is aligned with the guidelines of the VIVIFRAIL program [67], which specifically aims to prevent weakness and reduce the risk of falls.

In the hospital control and primary care cohorts, no specific intervention will be implemented.

In alignment with the VIVIFRAIL program guidelines, our intervention program is designed to prevent weakness and reduce the risk of falls. Patients will be informed about the program’s benefits during recruitment and follow-up. Within the supervised exercise program, interventions and patient progress will be personalized and monitored, taking into account functional capacity, comorbidities, and the individual’s previous experiences. The sessions will be closely monitored at the gym of the Geriatrics Department of the hospital. The individualized sessions will enable discussions on progress with the patient throughout the intervention, fostering motivation for retention and completing the follow-up. Participants will also be encouraged to adopt healthy habits, such as incorporating an additional 90-minute walk per week. In case participants require medical care that may interfere with the ongoing study procedures, they will be managed according to the clinical routine and subsequently excluded from the study. In these cases, the data collected during the baseline examination will be used for a cross-sectional clinical evaluation of frailty and sarcopenia. However, musculoskeletal changes following the multicomponent exercise program will not be assessed.

Each principal investigator at every data acquisition center will oversee the monitoring and follow-up of participants included in their respective cohorts under usual care. They will make decisions regarding patient exclusion and assignment to interventions, and monitor termination criteria as deemed necessary. Additionally, participation in an exercise program is generally associated with a low risk (–1%) of adverse events [68], most of which are typically low-grade responses to exercise, such as muscle soreness [69]. However, in the event of adverse events or serious adverse events, they will be promptly reported to the Clinical Research Ethics Committees.

Data analysis will be conducted independently of cohort recruitment and follow-up. The evaluation of data will take place at the Deusto Institute of Technology (Ultrasound Data Evaluation Center) and Biodonostia (Blood-based Assays Evaluation Center). Initially, only anonymized data, including ultrasound images, raw data, and blood-based assays, will be transmitted to the data evaluation centers. In the case of combination models, a training set of clinical variables will also be transferred to the data evaluation centers. Reserved data sets will be kept at the clinical acquisition centers for independent testing. This includes data from various sites and information obtained during follow-up examinations.

A population of 200 participants will be recruited, with 100 participants in the primary care cohort, 50 participants in the hospital exercise cohort, and 50 participants in the hospital control cohort (2:1:1 ratio). Patient selection within each acquisition site will be conducted based on the clinical agenda until the recruitment goals are met. Each participant will undergo 2 examinations (baseline and follow-up), resulting in a minimum total of 400 examinations.

An estimated distribution of 85% of patients with frailty to 15% of healthy participants is anticipated in the hospital cohorts, while a distribution of 15% of patients with frailty to 85% of healthy participants is expected in the primary care cohort. Consequently, an accrual of 139 robust participants and 86 patients with frailty is projected. It is assumed that up to 10% of the participants may be excluded in data evaluation as a result of failed measurements or logistical challenges in collecting all variables. In cases where follow-up measurements are not feasible for patients, the baseline measurement will still be included in the evaluation of the primary outcomes. With 80% power, the planned population will enable the detection of a biomarker medium effect size of E/S=0.39, considering a 2-sided α level of .05. For the secondary goal, our sample size of 100 provides 80% power to detect a biomarker with a medium effect size (E/S=0.284) at a 2-sided α level of .05. The sample size was calculated using RiskCalc (Moody’s Analytics) [70].

The list of clinical variables collected at baseline and follow-up assessments is included in Textbox 2.

The ultrasound equipment used in this study is the L7 HD3 portable linear scanner (Clarius Mobile Health Corp.). Specifically designed for point-of-care ultrasound examinations, the probe of this device boasts a frequency range of 4-13 MHz with a center frequency of 7 MHz. To facilitate seamless data transfer, anonymized digital ultrasound data are transmitted from the scanner through a customized Wi-Fi network to a smart device, enabling real-time B-mode navigation and data storage. The important feature of the scanner is an integrated image processing package designed to optimize B-mode musculoskeletal image quality. This technology enhances the clarity and quality of musculoskeletal imaging during examinations. Additionally, the scanner is equipped with a research package designed to capture raw beamformed backscattering ultrasound data after the beamformer. These raw data are presented in an in-phase/quadrature complex baseband representation, commonly referred to as IQ data. The IQ data serve as the foundation for the implementation of tailored quantitative ultrasound algorithms [71].

All examinations will be conducted within a depth range of 0-60 mm, using 50% of the maximum acoustic output offered by the scanner. During measurements, efforts will be made to minimize skin-probe compression to maintain acceptable image quality. Following each examination, the acquired data will be uploaded to a HIPAA (Health Insurance Portability and Accountability Act)-compliant cloud service, offered by the scanner’s manufacturer, enabling the centralization of data (Figure 3).

The midpoint of the thigh will be determined and marked as the half-distance between the superior border of the patella and the anterior-superior iliac crest. Using the femur as a guide in the transverse view, the midpoint of each thigh will be located, and the various components of the quadriceps muscles (rectus femoris, vastus intermedius, vastus medialis, and vastus lateralis) will be identified (Figure 4).

In each examination, a minimum of 12 coregistered B-scan and raw data frames will be obtained. This will include 3 transverse and 3 longitudinal views for both thighs. The longitudinal examination will be conducted in a plane where fasciculations are visible, ensuring comprehensive coverage and detailed assessment of the musculoskeletal structures.

During both basal and follow-up acquisitions, venipuncture will be performed to collect one 10-mL serum blood tube and two 5-mL ethylenediaminetetraacetic acid (EDTA) blood tubes. The serum sample will undergo centrifugation and will be stored in 4 aliquots. One of the EDTA samples will be promptly frozen and stored at –80°C, while the other will be subjected to centrifugation for plasma and buffy coat extraction. These samples will be stored at –80°C at the clinical sites until transportation, which will be carried out with dry ice to the Blood-Assay Evaluation Center for subsequent processing and storage (Figure 6).

The expression of previously described biomarkers, including vitamin D, lutein zeaxanthin, troponin T, pro-brain natriuretic peptide, soluble receptor for advanced glycation end products (sRAGE) [85], and microRNAs, along with those associated with relevant pathways related to frailty such as inflammation (interleukin 6) and senescence (p16INK4A and p21CIP [56]), will be assessed in cells obtained from a blood sample. This evaluation will be conducted in a subsample of patients in baseline conditions and after intervention at both the transcriptional and protein levels, as outlined in Table 2. Erythrocytes will be lysed using Buffer EL (Qiagen), and total RNA from leukocytes will be isolated using the miRNeasy Mini Kit (Qiagen). Initially, RNA samples will undergo purification with the RNeasy Kit (Qiagen). Subsequently, the RNA will be retro-transcribed, and gene expressions will be quantified through quantitative polymerase chain reaction (PCR) using specific primers or probes and the ABI Prism SDS 7300 Real-Time PCR System (Applied Biosystems). Expression levels will be normalized to the expression of the enzyme glyceraldehyde-3-phosphate dehydrogenase (GAPDH) [86]. Furthermore, protein levels will be assessed in serum samples using Quantikine enzyme-linked immunosorbent assays and Luminex, as previously outlined [87].

The study population will be characterized based on demographic and clinical attributes. Categorical data will be summarized with frequency and percentage representations. Continuous data will be described using mean, SD, median, and the 25th and 75th percentiles. Furthermore, the distribution of ultrasound parameters with respect to the primary endpoint will be depicted using graphical techniques. The correlation of quantitative biomarkers will be evaluated for each multifactorial clinical evaluation parameter using the Pearson correlation coefficient for continuous variables and the Spearman correlation coefficient for categorical variables [88]. Unpaired tests, such as the Student t test, will be used for comparing means of biomarkers among various acquired patient cohorts and for stratified patient subgroups based on the frailty scales outlined in Textbox 1. To manage the extensive array of ultrasound and blood biomarkers, a robust statistical analysis plan will be implemented. This plan will incorporate correction methods to address the risk of type I errors associated with multiple comparisons. Specifically, for the primary outcome concerning biomarker-clinical associations, we will use Lasso regression. This technique automates the selection of relevant variables, reducing the necessity for extensive corrections and improving interpretability. Paired statistics will be applied in the statistical analysis for longitudinal monitoring, encompassing both interventions and clinical follow-up. The monitoring period in hospital and primary care environments will differ to accommodate the distinct follow-up workflows in each setting. The duration of primary care follow-up aligns with established norms [89,90]. To account for differences in follow-up duration among cohorts over time, our data analysis will incorporate appropriate statistical methods, potentially enabling adjustments for these variations. For the secondary objective, the impact of physical intervention in 50 patients will be compared with 50 patients with no intervention. With 80% power, the planned population size will enable the detection of a biomarker with a medium effect size of E/S=0.284 at a 2-sided α level of .05.

In the context of multiparametric data science models, unsupervised learning techniques will be assessed to cluster patient populations based on biomarker expression. Moreover, supervised multi-omics models will be trained using a training set of clinical variables. Stratified cross-validation will be used to derive model performance statistics, ensuring separate patient data in the training and validation folds and maintaining a balanced distribution of health participants and those with frailty in both training and validation sets. Consecutively, sample imputation will be used to address missing data in the training set, while missing reference variables will be excluded during validation and testing. Sensitivity analyses will be conducted to evaluate the impact of the imputation method on the robustness of our results. Reserve data sets, comprising data from various sites and follow-up examinations, will be used for model testing.

Biomarker reproducibility will be evaluated using the interclass correlation coefficient and the Bland-Altmann method, taking into account repeated measurements within an examination. Subgroup analyses will be conducted to explore variations across patient subpopulations, and regression models will be used to examine the influence of covariates, thereby enhancing methodological precision.

This study protocol received approval from the Research Ethics Committee of Albacete (Spain) with reference CEIm-2021-51. The study will be conducted in adherence to the principles outlined in the Declaration of Helsinki [91]. Before participation, written informed consent will be obtained from all participants, and their data will be handled in accordance with HIPAA guidelines. Additionally, consent will be sought for the review of participants’ medical records and for the collection of blood samples to assess biomarkers. In the event of necessary modifications to this protocol during the clinical research, the changes will undergo review by the hospital ethics committee and will be implemented only after obtaining approval. The trial results will be communicated to participants via email by the investigators. Participants involved in this study will not receive any form of compensation for their participation.

The study information will be stored in Microsoft Excel 365 within an electronic database. This database will comprehensively capture individual data, including baseline characteristics, pre- and postassessment data, and any potential adverse events. Notably, no personally identifiable patient information, apart from the unique trial identification number, will be included in the metadata associated with the recorded ultrasound and biomarker data. Ultrasound and biomarker data will be uploaded to a HIPAA-compliant cloud service provided by the ultrasound system manufacturer, enabling centralized evaluation by team members. As outlined in the reference manuscript, upon completion of the study, digitized data sets encompassing ultrasound data, biomarkers, digitized blood-based biomarkers, and anonymized multifactorial clinical evaluations will be published in open-access repositories, such as Zenodo (CERN). Biological samples will be securely stored in the Basque Biomarker Center (Biobanco). Additionally, software models derived from ultrasound raw data will be made accessible through open-source software repositories (eg, GitLab).

The anticipated outcomes of this study are the identification of associations and the development of combination models between quantitative ultrasound and blood-based assays with multifactorial clinical assessments, specifically focusing on muscle quality and frailty. The use of noninvasive portable technologies facilitates the implementation of the clinical protocol in both hospital and primary care environments [92].

The primary findings concerning the correlations of ultrasound geometric variables with clinical and functional variables align with existing clinical literature [17,35,36,41]. These preliminary results are encouraging, demonstrating robust correlations between echo geometric parameters, measured in a standardized manner, and various clinical parameters. Notably, moderate correlations are observed between ultrasound geometrical parameters and muscle mass (DXA), while functional parameters and global scales of sarcopenia and frailty exhibit weak yet significant associations. These results serve as a foundational baseline for the development of advanced models based on raw data (Table 1) and blood biomarkers. It is anticipated that these advanced models will exhibit superior correlations with basal characteristics and enhanced sensitivity to changes in muscle quality.

Recent literature indicates that impaired mobility associated with aging is not solely attributed to changes in skeletal muscle mass; other factors contributing to muscle quality also play a crucial role. Notably, alterations in muscle tissue composition, characterized by elevated levels of intramuscular adipose tissue and intramyocellular lipids, have been identified as factors that negatively impact muscle functional capacity [17,65,93]. Our hypothesis suggests that quantitative ultrasound biomarkers derived from raw data may exhibit superior discriminative performance, greater reproducibility, and enhanced ease of use compared with the current state-of-the-art assessments based on B-mode images. B-mode morphological ultrasound parameters are primarily linked to muscle mass and demonstrate limited sensitivity in the diagnosis of sarcopenia. Quantitative ultrasound biomarkers derived from raw data have consistently demonstrated their ability to capture both tissue composition and microstructural properties. Furthermore, they have exhibited a greater capacity to encode richer information content compared with ultrasound B-mode images in artificial intelligence models, as highlighted in prior research [52,54,83,84]. Ultrasound spectroscopy parameters and tissue acoustic properties, such as speed of sound and attenuation, have been associated with tissue composition [47] and viscoelastic changes in muscle [49] among older individuals with sarcopenia. Specifically, speed of sound has demonstrated correlations with MRI adiposity estimates in the calf muscles [94], CT assessment of the psoas muscle [95], and short-term changes in muscle due to immobilization [96]. The ultrasound statistical analysis of the envelope signal in soft tissues has been correlated with the concentration, spacing, and directionality of microstructural scatterers [78]. Additionally, texture features derived from radiomics analysis capture musculoskeletal composition and microstructure. This approach has been successfully applied to differentiate various musculoskeletal conditions, including muscle spasticity, dynapenia, myositis, fibromyalgia, Duchenne muscular dystrophy, and exercise-induced muscle damage [55,97]. Being acquired at the early stages of ultrasound image formation, ultrasound raw data may play a crucial role in minimizing equipment- and operator-dependent bias. Additionally, they have the potential to provide data-based guidance for examiners with limited sonographic acquisition expertise.

Molecular biomarkers obtained from blood assays are linked to functional changes between healthy individuals and those with frailty. However, the absence of single established predictor biomarkers is primarily attributed to several factors. These include the heterogeneity and limitations of the scales or indices used to detect sarcopenia and frailty; variations in age, sex, and characteristics across different populations; small sample sizes; limited longitudinal clinical studies; a lack of characterization on interventions to assess their potential reversibility; and differences in techniques and cut-offs used for biomarker measurement. In this study, we incorporate a panel of biomarkers rather than assessing individual molecules. This approach is aimed at providing a more comprehensive reflection of the accumulation of damage associated with age-related syndromes [98]. Additionally, the potential of these biomarkers to assess reversibility, a critical characteristic of frailty, will be evaluated upon completion of an intervention based on an exercise program, followed by a 16-week follow-up.

Sarcopenia and frailty, although related, are distinct aging phenotypes. Notably, the prevalence of sarcopenia is higher among older individuals with frailty than the prevalence of frailty among those with sarcopenia [28-32]. As part of our exploratory outcomes, we will consider the coexistence of sarcopenia and frailty, redefining the frailty prognosis based on the presence or absence of sarcopenia.

To the best of our knowledge, this cohort study stands out as one of the few that integrates raw data ultrasound measurements with blood samples to extract noninvasive biomarkers of frailty and sarcopenia in older adults. Notably, this study represents the first of its kind to establish a connection between quantitative ultrasound and blood biomarkers with the cross-sectional evaluation of frailty and sarcopenia in both hospital and primary care environments. The utilization of point-of-care ultrasound devices presents an opportunity to introduce ultrasound quantitative technology across various clinical settings. Notably, this study marks the pioneering effort to combine quantitative ultrasound with blood-based biomarkers to evaluate musculoskeletal changes following multicomponent physical exercise programs.

The study does have certain limitations, notably the absence of access to a gold standard for muscle quality assessment. Established radiological techniques such as CT and MRI serve as references for muscle composition and microstructure but come with patient complications and are not widely accessible in the clinical environments under investigation. Instead, we rely on a multifactorial clinical assessment, patient stratification, and carefully controlled interventions to evaluate changes in muscle quality. The exploration of ultrasound technology is limited to backscattering ultrasound data derived from beamformed raw data. Other ultrasound quantitative technologies, such as shear wave elastography [43] and blood flow measurements based on Doppler sequences [99], are excluded from the scope because of their lack of routine availability in point-of-care ultrasound devices and the added complexity they introduce to the protocol execution. The disparate monitoring periods in hospital (16 weeks) and primary care environments (1 year) are designed to accommodate the distinct follow-up workflows in both settings. We also acknowledge that the randomization of populations in the secondary goal is conducted on an institutional basis, which could introduce statistical bias if the populations in the 2 institutions differ. To address this potential bias, we will perform a post hoc analysis to examine the characteristics of the populations in the various recruitment centers. The dimensioning of the primary care follow-up aligns with periods established in previous studies for the general population [89,90].

In summary, the outlined study protocol aims to integrate portable technologies, leveraging quantitative muscle ultrasound and blood-based biomarkers, for an objective cross-sectional assessment of muscle quality in both hospital and primary care environments. The study endeavors to yield data that will facilitate the exploration of associations between biomarker combinations and the cross-sectional clinical evaluation of frailty and sarcopenia, alongside an examination of musculoskeletal changes following multicomponent physical exercise programs.