Fourier-Transform Infrared Spectroscopy as a Screening Tool for Osteosarcopenia in Community-Dwelling Older Women

Abstract

Osteosarcopenia is a complex geriatric syndrome characterized by the presence of both sarcopenia and osteopenia/osteoporosis. This condition increases rates of disability, falls, fractures, mortality, and mobility impairments in older adults. The purpose of this study was to analyze the Fourier-transform infrared (FTIR) spectroscopy diagnostic power for osteosarcopenia in community-dwelling older women (n = 64; 32 osteosarcopenic and 32 non-osteosarcopenia). FTIR is a fast and reproducible technique highly sensitive to biological tissues, and a mathematical model was created using multivariate classification techniques that denoted the graphic spectra of the molecular groups. Genetic algorithm and support vector machine regression (GA–SVM) was the most feasible model, achieving 80.0% of accuracy. GA–SVM identified 15 wave numbers responsible for class differentiation, in which several amino acids (responsible for the proper activation of the mammalian target of rapamycin) and hydroxyapatite (an inorganic bone component) were observed. Imaging tests and low availability of instruments that allow the observation of osteosarcopenia involve high health costs for patients and restrictive indications. Therefore, FTIR can be used to diagnose osteosarcopenia due to its efficiency and low cost and to enable early detection in geriatric services, contributing to advances in science and technology that are potential “conventional” methods in the future.

The aging process is related to several biological changes, such as musculoskeletal degeneration, which can be exacerbated by the presence of multiple morbidities and a sedentary lifestyle that affect older adults. These conditions reduce functional capacity and increase mortality rates in this population (1). One of the musculoskeletal disorders that is most related to the aging process is sarcopenia, which is characterized by reduced muscle strength associated with loss of muscle mass and impaired functional performance (2). Sarcopenia may affect around 6.8% (3) of people aged 65 years or more and be present in 30% of those who are more than 80 years old (4). It can increase the risk of becoming functionally dependent and disabled, falling and having a fracture, and death (5). Sarcopenia is also related to another musculoskeletal disorder that is strongly associated with aging, namely osteoporosis, which is characterized by loss of bone mass and damage to its microarchitecture (6). Osteoporotic fractures are followed by a substantial economic impact due to their high-cost treatment that demands extended follow-up (7).

Due to the complex and important muscle and bone cross-talk, a new geriatric syndrome has been reported recently. Osteosarcopenia is characterized by the presence of both sarcopenia and osteopenia (t score of <−1) or osteoporosis (t score of <−2.5). This condition is related to risk factors such as sex (female), oral glucocorticosteroid intake, poor calcium-based diet, hyperparathyroidism, and rheumatoid arthritis (6). Some of the other factors that have been associated with osteosarcopenia are changes in the expression of vitamin D receptors (8) and genetic factors that can influence peak bone mass and muscle strength (9). Lifestyle habits also influence bone and muscle health, as is the case of excessive alcohol consumption, directly affecting osteoblast function (10) and being associated with reduced muscle mass (11). It is also known that older women are more likely to develop both osteoporosis and sarcopenia (12). Osteosarcopenic older adults also present a higher prevalence of disability, falls, fractures (13), and mortality risk (14). Approximately 37% of the older community population with a history of previous falls has been reported to have osteosarcopenia (15). Moreover, these people are more inclined to present mobility impairments and have a higher prevalence of nontraumatic falls (15).

Several techniques with different costs, feasibility, and ease are available to assess and diagnose sarcopenia. The European Working Group on Sarcopenia in Older People (EWGSOP) recommends handgrip strength as being the easiest and cheapest technique to assess muscle strength due to the wide correlations with important outcomes in the geriatric and gerontology field found in the literature (2). Dual-energy x-ray absorptiometry (DXA) is known as a noninvasive, fast, and reliable method to measure body composition, which precisely distinguishes the tissues and provides information regarding both muscle and bone mass (16). However, the equipment is expensive and is not portable, which could hinder its use in epidemiological studies.

Due to the rapid growth of the aging population, there is an increasing need to develop strategies to prevent diseases and disabilities associated with aging. Therefore, there is a necessity to find faster and more effective ways to measure the impact of aging in order to reduce them. An alternative that has been explored in biological studies and presents promising results is biospectroscopy (17). Fourier-transform infrared (FTIR) spectroscopy is a fast and reproducible technique that requires small sample amounts and is highly sensitive to the molecular structure of components of biological tissues such as lipids, proteins, and nucleic acids (18). This technique has shown promising findings in several studies with different populations, such as ovarian cancer (19), Alzheimer’s disease (20), and fibromyalgia (21).

Therefore, the objective of this study was to analyze the diagnostic power of FTIR spectroscopy to differentiate older women from the community with and without osteosarcopenia through serum.

Method

Design and Participants

This is a cross-sectional diagnostic study approved by the Ethics Committee of the Federal University of Rio Grande do Norte for Research, under number 2.368.206. Methods were carried out in agreement with the ethics guidelines. The sample was composed of community-dwelling women who were able to walk independently for at least 400 m with or without walking aids; did not have cognitive impairments (Cognitive Leganés Task >22); did not have a history of cancer in the previous 5 years; did not have acute inflammatory or immunological conditions, such as rheumatoid arthritis or systemic lupus erythematosus; did not have orthopedic or neurologic disabilities that would interfere in the test results; did not practice physical activity regularly (at least 3 times per week); did not take immunosuppressive and/or corticoid drugs in the previous 3 months. All participants gave their informed and signed consent. A wide dissemination of the research project was carried out for the recruitment process of participants, in addition to an active search in groups for older people. In total, 191 older women were evaluated and 167 older adults met the eligibility criteria and agreed to participate in the study.

Procedures

Assessments were carried out at 2 different moments: an interview with participants through a questionnaire with sociodemographic and general health status data; and a cognitive and physical function assessment (muscle strength and physical performance).

Physical performance.—A JAMAR hydraulic manual dynamometer (Sahen) was used to measure handgrip strength (kg). Participants were evaluated in a sitting position with both feet touching the ground, with 90° of flexion in the hips, knee, and elbows. The wrists remained in a neutral position while their elbows rested on the armchair. The participants were then asked to perform 3 tests in the dominant hand, and the mean value was recorded for further analysis. The cutoff point adopted to determine weakness was <16 kg as determined by the revised EWGSOP (2).

Gait speed was also measured. Participants were asked to walk at their usual speed for 4 m, and the evaluator recorded the time. Low physical performance was identified when the gait speed was <0.8 m/s (2).

Body composition and osteosarcopenia diagnosis.—Body composition and anthropometric analysis were conducted on the same morning by a trained physiotherapist. The participants were instructed to be fasting for at least 4 hours and to take their daily medicines after the assessments. DXA was used to estimate muscle and bone mass, where the participant was asked to lie down and remain immobile for 15–20 minutes. They were also instructed to wear comfortable clothing without metallic props. Reduced muscle mass was observed when the appendicular skeletal mass/height (2) was less than 5.5 kg/m².

DXA was also used to measure bone mineral density (BMD; femoral neck). The cutoff points adopted were the recommendations of the International Society for Clinical Densitometry (22) and the Brazilian Society for Clinical Densitometry (23). Values of t scores up to −1.0 DP are considered normal; −1.01 to −2.49 DP represents osteopenia; and t scores equal to or less than −2.5 DP characterize osteoporosis. Sarcopenia was diagnosed according to the EWGSOP criteria (low muscle strength and low muscle mass), and this condition was considered severe when low physical performance was observed (2). The participants received an osteosarcopenia diagnosis when the participant presented both sarcopenia and reduced BMD (osteopenia and/or osteoporosis).

Storage and preparation of serum samples.—Blood samples were obtained and analyzed in the same laboratory and by the same technician. They were centrifuged at 4 000 rpm per 10 minutes to obtain the serum, which was stored in aliquots at −80°C for posterior analysis. All samples were defrosted at ambient temperature for 30–40 minutes before spectrometric analysis, and the protein precipitation process was conducted. Precipitation was performed by adding 1.5 μ of perchloric acid at 7M to a 100 µ aliquot of serum. The aliquot was placed in a vortex (FlexVortex 2, Loccus) for 15 seconds, and centrifuged for 12 minutes at 12 000 rpm and 4°C. Precipitation was performed to obtain better results with biomolecules in the plasma known as potential osteopenia and/or osteoporosis biomarkers (proteins).

Fourier-transform infrared spectroscopy.—Precipitate samples were analyzed through a SHIMADZU model IRAffinity-1 spectrometer equipped with an attenuated total reflection (ATR) accessory. The spectra were obtained at a resolution of 4 cm⁻¹, 34 seconds (32 scans) per spectrum. Air was used as the background for all samples. An amount of precipitate mass was placed at the ATR crystal, and a range was obtained. The crystal was washed with alcohol (70% v/v) and another amount of precipitate mass was analyzed. Then, a mathematical model was created using multivariate classification techniques that denoted the graphic spectra of the molecular groups.

Statistical Analysis

Principal component analysis

Principal component analysis is an experimental technique that reduces the original data by a few numbers of orthogonal variables to each other and has the ability to explain most of the original information, meaning without decreasing the variance (24). These variables are called principal components (PCs). They are arranged in such a way that the first PC explains most of the original information and then the other PCs subsequently explain a smaller part of the information that has not yet been explained. Thus, the matrix containing the sample data that may or may not contain information about the sample classes is decomposed as follows:

On the other hand, the successive projection algorithm (SPA) is a method of selecting variables with great potential for applications in biological data. It is an advanced variable selection method that seeks to solve the problems related to collinearity, in which the variables whose information content is essential are selected through a process generated through a series of interactions starting with a variable, incorporating new ones for each interaction until a certain number of variables are chosen. There are several advantages to using this model because it has a deterministic nature and maintains the same variable space; thus, the selected variables have the same physical meaning in relation to the original data, keeping the original information. The optimal number of variables for SPA was determined from the minimum cost function G calculated for a given set of internal validation data.

where $gn$ is defined as

In turn, the LDAmodel is a boundary-discriminating technique that aims to find limits that separate the classes of samples. This technique seeks to obtain linear limits that separate the 2 groups, assuming a combined covariance matrix over all classes.

To obtain the discriminant profile, the classification score linear discriminant analysis (LDA) (⁠$Lij$⁠) is calculated for a given class k by the following equation, considering that the covariance matrices of each class are considered equal:

Support vector machine

The support vector machine (SVM) classifying technique was also used in this study, which is a supervised method of a nonlinear classificatory nature with high precision. It is a model considered to be of high complexity and it is necessary to optimize the kernel function and determine the SVM parameters (25). The radial base function (RBF) was used. This classifier is used to transfer data to a resource space using a nonlinear discriminating criterion, after which a linear decision is then constructed in the resource space in order to separate the classes from the analyzed samples (26). The RBF is calculated as follows:

where $xi$ and $zj$ are sample measurement vectors and $γ$ is an adjustment parameter that controls the width of the RBF. For this case, this parameter was set to 1. The SVM classification rule is obtained by the following equation:

where $NSV$ is the number of support vectors; α _i is the Lagrange multiplier; $yi$ is the affiliation to the respective class (±1); $k(xi,zj)$ is the kernel function; and b is the polarization parameter. These SVM parameters were obtained from the validation set.

For internal validation, leave-one-out cross-validation (LOO-CV) was applied, being the latter implemented with fourfold (K = 4) and 10 repeats (number of internal validations: fourfolds’ 10 repeats = 40 validation runs). At each procedure, 25% of the dataset was randomly selected for validation purposes, and the remaining 75% was used for training setting.

Figures of merit

The statistical parameters for evaluating the classification models were accuracy (AC), sensitivity (SENS), specificity (SPEC), F-score, and G-score. The AC refers to the percentage of samples correctly classified by the model. SENS measures the proportion of correctly identified positive results, while SPEC measures the proportion of correctly identified negative results. Thus, sensitivity can be understood as the probability of finding a positive result when the disease is present, and specificity can be interpreted as the probability of finding a negative result when there is a disease. The F-score represents the weighted average of sensitivity and specificity; and the G-score represents an account for the model’s accuracy and sensitivity without the influence of positive and negative class sizes (27). These parameters are calculated based on the equations presented in Table 1.

Results

Our sample comprised 32 osteosarcopenic women. The same amount of non-osteosarcopenic women was randomly selected as controls. The mean age was 73.5 (±6.8) years. Low BMD was observed in 92.2% of the sample. Among sarcopenic older women, 62.5% had severe sarcopenia, with reduced gait speed associated with reduced handgrip strength and muscle mass. Mean t scores were −1.1 (±1.00) for non-osteosarcopenic (range: −2.9 to −1.1) and −1.8 (±1.2) for osteosarcopenic older women (range: −3.3 to 1.5). Sociodemographic and clinical data are available in Table 2.

FTIR spectral data in the fingerprint region (900–1800 cm⁻¹) were preprocessed through Savitzky–Golay smoothing followed by automatic weighted least squares baseline correction and vector normalization in order to avoid nonbiological interference that could impair the results. A training set composed of 70% of samples (22 negative and 22 positive) and 30% of samples (10 positive and 10 negative) was used for test models.

Raw absorbance spectra are shown in Figure 1A and C. The preprocessed spectra (Figure 1B and D) were used to obtain several classification techniques (Table 3). The most feasible model was GA–SVM, achieving 80.0% accuracy, 70% sensitivity, and 90% specificity.

The wave numbers were selected by GA–SVM as responsible for class differentiation: 952, 975, 995, 1134, 1165, 1246, 1247, 1309, 1317, 1477, 1500, 1517, 1533, 1556, and 1579 (Figure 2; Supplementary Table 1).

Discussion

Osteosarcopenia has gained notoriety in recent years, but few studies have been conducted in older populations. This study is the first to use an infrared diagnostic method, and aimed to determine if infrared spectroscopy is suitable for screening osteosarcopenia in older women. The values of precision, specificity, and sensitivity were suitable for identifying older women with osteosarcopenia.

FTIR is an excellent method for biological analysis. FTIR is a type of high-energy vibrational spectroscopy performed in a field with a wave length between 750 and 2500 nm, whose reflectance spectrum will be affected by the interaction of radiation with the analyzed samples. The most important spectral regions measured by FTIR are nucleic acids (asymmetric $PO2−$ in DNA and RNA at ~1.225 cm⁻¹, carbohydrates (C–O stretching at ~1.155 cm⁻¹), proteins (amide II at ~1.550 cm⁻¹ and amide I at ~1.660 cm⁻¹), and lipids (C=C stretching at ~1.750 cm⁻¹ (28).

This study innovated using solid biological samples for FTIR analysis in the context of osteosarcopenia. Water acts as an interference factor in the analysis of liquid samples, which can compromise the sensitivity and detection limit of the technique (29). The precipitation of biomolecules from blood plasma therefore minimized water interference during infrared analysis. Many methods are used to classify spectral data, and choosing the most suitable one depends on the characteristics of the samples studied. The SVM method is part of the new generation of algorithms used for classification and regression tasks. Furthermore, one of the main objectives of the SVM model is that it can operate with a nonlinear model, allowing good generalization performance even with small datasets (30). The reduced data and the selection of variables can facilitate the computational analysis of the SVM, as is the case of the use of GA (31), which reduces the data in an evolutionary process where a series of more appropriate data from variables are selected, but the original data dimension is maintained (32).

A total of 15 wave numbers were defined by the model, differentiating control cases as pointed out by standard values that were used to determine the wave numbers (33). These include carbohydrates, amino acids, lipids, and nucleic acids. Skeletal muscle comprises 5% carbohydrates and lipids (34). High-carbohydrate diets have been associated with greater muscle hypertrophy (35) and have been indicated to maintain high glycogen levels to maintain muscle contractions during high-intensity strength training and improve muscle adaptation and recovery by increasing protein synthesis and suppressing protein breakdown between workouts (36). Furthermore, it plays an important role in body composition, as it tends to reduce the amount of fat mass (37).

Amino acids were found in several wave numbers in our selected model, and are pointed out together with growth factors as the most important signals for the activation of the mechanistic (or mammalian) target of rapamycin (mTOR) as it is not properly activated in the absence of amino acids (38). mTOR is a kinase that regulates cellular functions associated with the promotion of cell growth and metabolism and is part of 2 protein complexes: mTOR1 complex (mTORC1) and mTOR2 complex (mTORC2), which play a fundamental role in the coordination of anabolic and catabolic in response to growth factors and nutrients (39). When activated, mTORC1 activates cell growth and proliferation, promoting protein synthesis, biogenesis, and lipid metabolism, in addition to reducing autophagy.

Hydroxyapatite (HA), as evidenced in the 952 cm⁻¹ wave, is an inorganic bone component whose porous structure serves as a substitute for rigid tissues and helps osteoconductivity (facilitator of osteoblast differentiation) and osteogenesis (bone production process) (40). Recent studies have also shown the presence of HA related to signal peptides for bone regeneration (41). Low BMD is characterized by microarchitectural deterioration and loss of bone tissue and mass that increases the risk of fracture (42), affecting more than 200 million people worldwide (43). A high number of osteopenic/osteoporotic women were included in our study, regardless of the presence of sarcopenia. The prevalence of osteoporosis among postmenopausal women is approximately twofold higher than among men at the same age, reaching 6%–33.2% of the Brazilian population (44). The prevalence of osteopenia is even higher (~45%) (45).

Osteoporosis emerges as a public health concern as the world’s aging population rapidly increases, and may lead to greater morbidity, mortality, and socioeconomic costs (46). A study conducted in Australia found that the expenses needed to treat the consequences of osteoporosis increased 3 times in 10 years, providing strong public health attention to promote bone health to reduce future costs (47). Public expenses with drugs for bone structure and mineralization correspond to approximately 10.9% in Brazil (48). It is still important to note that more than half of the sample studied was diagnosed with severe sarcopenia according to the EWGSOP2 criteria. Severe sarcopenia is even more worrisome, affecting up to 5.6% of community-dwelling older adults (49). This health condition is associated with a higher number of deaths, which highlights the report that physical performance plays a vital role in reducing mortality in this same population (50).

In view of the complications resulting from osteosarcopenia, access to early detection of this musculoskeletal disorder is essential to avoid more serious complications, such as fractures and death. Even so, Brazil still has a demand for imaging tests and low availability of instruments that allow the observation of these pathologies, which involves high health costs for patients and restrictive indications. Therefore, FTIR can be used to diagnose osteosarcopenia due to its efficiency, low cost, and to enable early detection of this health condition in several geriatric services, which contributes to advances in science and technology that are potential “conventional” methods in the future.

In conclusion, our findings indicate that FTIR is a good diagnostic method in osteosarcopenic older women in the community, with the GA–SVM model being the most appropriate. Therefore, it is suggested that this technique can be used to detect osteosarcopenia early in this population, minimizing the deleterious effects of this health condition when installed in the long term. However, studies with larger samples and including both genders are necessary to verify the applicability of the FTIR on a large scale and considering the different variability of the population studied.

Funding

This work was supported by Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq—grant 408358/2016-5).

Conflict of Interest

The authors declare that they have no conflict of interest.

Author Contributions

R.V.M.F. designed and conducted experimental work, performed data collection, analyzed the data, and wrote the manuscript; D.L.D.F. and T.G.S. were responsible for multivariate analysis and the construction of the models; I.R.D.O., C.S.G., and P.M.S.D. were enrolled on conceptualization and data collection; G.C.B.G. and K.M.G.L. provided chemometric expertise; G.D. and G.C.B.G. provided pathological expertise and clinical insight; R.O.G. designed, analyzed data, and finalized the manuscript.

Data Availability

The datasets generated and/or analyzed during the current study are available in the ZENODO repository, https://zenodo.org/record/6512453#.YnBEdNrMJPZ.

References