Study Design and Participants
This multicenter, observational, analytical, cross-sectional, quantitative study analyzed data from the National Academic Cooperation Program (PROCAD), an initiative funded by the Brazilian Coordination for the Improvement of Higher Education Personnel (CAPES).
The study sample population comprised 700 adults aged ³80 years and was obtained from the PROCAD electronic database. Participants were recruited from various settings, including family homes, long-term care facilities, and outpatient geriatric clinics in three Brazilian cities: Brasília (n=196), Passo Fundo (n=272), and Campinas (n=232).
Ethical approval
This study was approved by the Institutional Ethics Review Boards (IRBs) of the three participating institutions, ensuring adherence to research norms for human subjects defined by Brazil’s National Research Council. Written informed consent was obtained from all study participants prior to participation. Specific study approval was granted by the following university IRBs: Universidade de Passo Fundo 2.097.27/2017, Universidade Católica de Brasília 1.290.368/2015, and Universidade Estadual de Campinas 3.061.534/2018.
Eligibility criteria and clinical outcomes
Patients were included if they had adequate vision, hearing, and cognitive capacity to participate in the study. Cognition was assessed using the Mini Mental State Examination (MMSE) to ensure adequate temporal orientation, spatial orientation, immediate memory, command comprehension, and reading ability[9-12]. Patients were excluded if they were unable to (i) stand/walk or (ii) respond to questions due to conditions such as lower limb amputations, hemiplegia, stroke sequelae, or other neurologic conditions that impaired mobility or verbal communication.
Study outcomes
The primary outcome was cognitive impairment, defined using prespecified MMSE cutoffs [9]. The secondary outcome was functional dependence, defined as the need for total or partial assistance ³for 5 instrumental activities of daily living (IADL)[12].
Sociodemographic and clinical assessment
In-person interviews conducted by trained staff assessed sociodemographic and clinical variables using standardized questionnaires[12,13]. Anthropometric measurements included weight, height, body mass index (BMI), and waist circumference (WC). Weight was measured using a 300 kg-capacity digital electronic scale with a stadiometer (Welmy® W300 brand). Height was measured after a deep breath, with the patient remaining in a completely erect position. BMI was calculated as weight (kg) divided by height squared (m²). Standardized cutoffs were used to categorize BMI based on World Health Organization (WHO) guidelines as underweight (BMI <18.5 kg/m²), eutrophic (BMI 18.5-25 kg/m²), overweight (BMI>25.0 to <30.0 kg/m²), or obese (≥30.0 kg/m2)[13]. WC was measured at the midpoint between the iliac crest and the last rib using an inelastic tape. Central adiposity was determined based on WC cutoffs associated with elevated cardiometabolic risk (≥88 cm in women and ≥102 cm in men)[13].
To assess sarcopenia, we used the EWGSOP2 criteria[14] and considered gait speed ≤0.8 m/s, grip strength <27 kg and/or appendicular lean mass/ht2 ≤7.0 kg/m2. Insomnia severity was assessed using the 7-question Insomnia Severity Index (ISI). Scores of 0-7 indicate no clinically significant insomnia, 8-14 indicate subthreshold insomnia, 15-21 indicate moderate insomnia, and 22-28 indicate severe insomnia[16].
Cognitive status was evaluated using the 30-point Mini Mental State Examination (MMSE) [17]. Cutoff scores for cognitive impairment were based on the Brazilian College of Neurology guidelines [9], which stratify cutoffs according to education level: £17 for illiterate individuals; £22 for those with 1-4 years of schooling; £24 for those with 5-8 years of schooling; and £26 for those with ³9 years of schooling. Using education-adjusted norms allows the MMSE to accurately screen for cognitive dysfunction across varying sociodemographic profiles.
Biomarker analyses
Biochemical and inflammatory marker levels and DNA methylation were assessed in a subset of 175 participants from the Brasília site. Blood samples collected into EDTA tubes were divided into aliquots of less than 1 mL for DNA methylation assays. Additional blood samples were centrifuged at 2,500 rpm for 15 minutes at 25°C to evaluate inflammatory markers. All samples were obtained by trained nursing staff through venipuncture using a vacuum system and stored at –20°C until further analysis.
Standard clinical laboratory protocols quantify fasting glucose, triglycerides, total cholesterol, and lipid fractions through enzymatic, kinetic, or colorimetric tests (as appropriate) using reagents compatible with HumanStar 600 equipment (InVitro®). LDL-C was estimated using the Martin–Hopkins formula. Ultrasensitive C-reactive protein and glycated hemoglobin were quantified using turbidimetry and ion-exchange high-performance liquid chromatography, respectively.
A cytokine panel was used to evaluate the serum interferon-gamma, interleukin-2, -4, -6, and -10 concentrations using a high-throughput flow cytometry-based bead assay kit (FACS Verse model; BD Biosciences, USA) according to the manufacturer’s protocol (BD Biosciences). Titration curves generated from kit-provided standards facilitated quantitative analysis. All the scores were estimated through curve interpolation. In cases where a sample yielded outlying readings beyond the expected range, we conducted additional assays using either the original or diluted samples (as necessary) until we obtained a minimum of 300 events for each cytokine bead used. Subsequently, we analyzed all the data using FCAP software, version 3.0 (BD Biosciences).
DNA extraction was carried out using a commercial kit (QIAamp® DNA Blood Mini Kit, QIAGEN, Germany) in accordance with the manufacturer’s guidelines. Quantification by a NanoDrop Lite spectrophotometer (Thermo Fisher Scientific®, USA) preceded bisulfite conversion, and analysis of methylated cytosines (5-mC) was performed through a colorimetric ELISA-based assay (MethylFlash® Methylated DNA Quantification Kit, Epigentek, USA). The results, expressed as the percentage of methylated DNA (5-mC) in the total DNA sample, were extrapolated from a standard curve generated using serial dilutions of the synthetically methylated positive control DNA provided in the kit. The optical density (OD) of the curve was used to determine the 5-mC concentration.
Data analysis and statistical methods
Multiple imputations were performed using predictive mean matching (for numeric variables), logistic regression (for binary variables with 2 levels), and Bayesian polytomous regression (for factor variables ≥2 levels), and missing values (MVs) were addressed for selected covariates in the cohort comprising 493 participants using the mouse package for R. Imputed values, residual distributions and convergence coefficients were checked. Missingness rates reached 35% (mean 14%) for some variables, but multiple imputations were not performed for the main exposures (insomnia and sarcopenia), outcomes (cognitive and functional decline), or biomarkers. Multiple imputation procedures resulted in five complete data sets, each containing different estimates of missing values for all 493 cohort participants. The complete imputed data sets were pooled and merged into a single data set for downstream statistical analyses.
The assumptions needed for univariate statistical analyses were rigorously assessed, including checking for identifying atypical data points and evaluating the distribution of variables. For descriptive analyses, categorical variables are presented as the frequency (f) and percentage (%), while continuous variables are summarized as either the mean±SD or median (Md) and interquartile range (IQR), depending on the distribution of the variables.
Participants were stratified into four study subgroups according to the occurrence of sarcopenia, insomnia, both, or none of the conditions. ANOVA followed by post hoc analyses involving multiple comparisons of group means were employed to compare clinical characteristics among the study subgroups. When significant differences emerged, the Bonferroni correction was applied to control for type l errors. Nonnormal data were subjected to post hoc analysis based on their rank. The chi-square (ꭕ²) test was used for comparing categorical variables. Whenever the minimum number of cases per cell was not met, Fisher's test correction factor or the likelihood ratio was applied, as appropriate.
Forward stepwise logistic regression models were used to identify predictors of cognitive impairment (primary outcome) and functional dependence (secondary outcome). The candidate variables included all the sociodemographic, clinical and biochemical characteristics of the participants. Isolated or combined indicators for insomnia, sarcopenia or both conditions were forcibly entered to quantify isolated or combined effects on outcomes. Wald test p values <0.05 were considered to indicate statistical significance and were subsequently used to select variables for the models. Insomnia, sarcopenia, and both conditions were included in the models, irrespective of their significance according to the Wald test. The final models excluded contributors to outcomes (variables used to diagnose cognitive impairment or functional dependence) and covariates exhibiting high collinearity (variance inflation factor >5). Statistical analyses were carried out with R Studio v.2023.03 for Mac.