This study was a prospective observational analysis nested within the framework of the PREDIMED-Plus trial (Fig. 1) [18]. Briefly, PREDIMED-Plus is a multicenter, parallel-group, randomized, single-blind clinical trial designed to evaluate the long-term effects of a comprehensive lifestyle intervention including an energy-reduced Mediterranean diet, physical activity promotion, and behavioral support for weight-loss (intervention group), compared with general recommendations to follow an ad libitum Mediterranean diet (control group). The main objectives of the trial are the primary prevention of cardiovascular disease and long-term body weight-loss maintenance. Details of the PREDIMED-Plus protocol have been published elsewhere [19, 20], and are available online at https://www.predimedplus.com/. Ethical approval was obtained from the institutional review boards of all participating centers, and all participants provided written informed consent. The trial was registered in 2014 at ISRCTN.com (ISRCTN89898870).

Study design. Abbreviations: ALM, appendicular lean mass; ALMI, appendicular lean mass index; DXA, dual-energy x-ray absorptiometry; MetS, metabolic syndrome; PREDIMED-Plus, Prevención con Dieta Mediterránea–Plus

Participants were community-dwelling adults aged 55–75 years with overweight or obesity (body mass index (BMI) 27 to 40 kg/m2) and metabolic syndrome (Fig. 1A) [21]. Metabolic syndrome was identified according to harmonized criteria established by the International Diabetes Federation, and the American Heart, Lung, and Blood Institute, requiring the presence of at least three of the following components: elevated blood pressure (≥ 130/85 mmHg), elevated plasma triglycerides (≥ 150 mg/dL), low plasma high-density lipoprotein cholesterol (HDL [< 40 mg/dL for males or < 50 mg/dL for females]), elevated fasting blood glucose (≥ 100 mg/dL), and central obesity (waist circumference ≥ 102 cm for males or ≥ 88 cm for females) [21]. Exclusion criteria included the following: (I) unwillingness to give written informed consent; (II) institutionalization; (III) pre-existing cardiovascular disease, psychiatric disorders, or bowel diseases; (IV) use of weight loss medication; and (V) inability to adhere to the intervention due to religious, dietary, or medical restrictions among others. Between October 2013 and December 2016, a total of 6874 eligible participants were randomly assigned in a 1:1 ratio to either the intervention group or the control group. Randomization was blinded to all research staff and principal investigators. In cases where two participants from the same household enrolled, randomization was performed by cluster, considering the household as the unit of allocation. For the purposes of the present study, we used data from a subsample of participants who underwent dual-energy X-ray absorptiometry (DXA) measurements in 7 of the 23 recruiting centers having access to DXA scanners (DXA sub-study). Among 2,519 eligible participants in this sub-study, 2 of the 7 recruiting centers having access to DXA scanners had available baseline measurements of TL in addition to DXA assessments (Fig. 2).

Flow chart of the study population. Abbreviations: DXA, dual-energy x-ray absorptiometry; PREDIMED-Plus, Prevención con Dieta Mediterránea–Plus

Blood samples were obtained after an overnight fast. Buffy coat fractions were isolated and stored at −80 °C until DNA extraction. Genomic DNA was extracted from frozen buffy coat samples using the Maxwell RSC Blood DNA Kit (Promega, Madison, WI, USA) on the Maxwell RSC Instrument, following the manufacturer’s instructions. Relative leukocyte TL was determined at baseline using a monochrome multiplex real-time quantitative PCR (MMqPCR) assay based on the method described by Cawthon [22]. In each reaction, telomere repeat copy number (T) and a single-copy reference gene (albumin; S) were amplified simultaneously. All samples were analyzed in triplicate in 384-well plates using a CFX384 Touch Real-Time PCR Detection System (Bio-Rad, Hercules, CA, USA).

PCR reactions were prepared using the QuantiTect SYBR Green PCR Kit (Qiagen, Hilden, Germany) in a final volume of 10 µL containing 10 ng of genomic DNA, telomere and albumin primers, and nuclease-free water. Primer sequences were as follows: telg, 5′-ACACTAAGGTTTGGGTTTGGGTTTGGGTTTGGGTTAGTGT-3′; telc, 5′-TGTTAGGTATCCCTATCCCTATCCCTATCCCTATCCCTAACA-3′; albu, 5′-CGGCGGCGGGCGGCGCGGGCTGGGCGGAAATGCTGCACAGAATCCTTG-3′; and albd, 5′-GCCCGGCCCGCCGCGCCCGTCCCGCCGGAAAAGCATGGTCGCCTGTT-3′. Each plate included a seven-point standard curve generated from reference DNA using a twofold serial dilution (150 to 2.34 ng/mL), negative controls, and calibrator samples to monitor inter-plate variability. Assay performance was assessed according to predefined criteria, including standard-curve linearity, amplification efficiency, and replicate agreement. Intra-assay and inter-assay coefficients of variation were calculated, and samples not meeting quality thresholds were repeated. Only runs with standard curve linearity R2 > 0.99 were accepted. TL was calculated as the telomere-to-single-copy gene (T/S) ratio using the standard curve method for relative quantification.

TL was measured in a subsample of 185 participants from the Reus and Pamplona centers of the PREDIMED-Plus study at baseline, based on prior experience in TL assessment in human samples and laboratory resource availability. These two centers were the only sites participating in this sub study (Fig. 2).

At baseline, and after 1, 3, and 6 years of follow-up, height (cm) was measured in light clothing and without shoes using a calibrated scale and wall-mounted stadiometer. Body composition was assessed by dual-energy X-ray absorptiometry using Lunar iDXA or DXA Lunar Prodigy Primo devices (GE Healthcare, Madison, WI). Scans were performed by trained radiology technicians following standardized manufacturer protocols for participant positioning. Each DXA device was calibrated daily according to manufacturer guidelines. Total body weight and regionally distributed lean mass in arms and legs were obtained. ALM (kg) was calculated as the sum of lean mass from the four limbs. BMI was calculated as (total body weight [kg]/height [m]2). The following ALMI was subsequently computed: the ALM to height ratio (ALM [kg]/height2 [m2]), the ALM to total body weight ratio (ALM [kg]/total body weight [kg] × 100), and the ALM to BMI ratio (ALM [kg]/BMI [kg/m2] × 100) (Fig. 1C).

Baseline covariates included sociodemographic and lifestyle information. Data on age, sex, education level, marital, and smoking status were collected through administered questionnaires. Physical activity was estimated using the validated REGICOR (Registre Gironí del Cor) Short Physical Activity Questionnaire for adults adapted from the Minnesota Leisure Time Physical Activity Questionnaire (MLTPAQ) [23], whereas sedentary behavior was assessed using the validated Spanish version of the Nurses’ Health Study questionnaire [24]. Medical history, including type 2 diabetes, hypercholesterolemia, and individual components of the metabolic syndrome (hypertension, hypertriglyceridemia, low-HDL cholesterol, hyperglycemia, and central obesity), as well as medication use, was obtained from self-reports or extracted from medical records. Habitual dietary intake was assessed at baseline using a validated 143-item semi-quantitative food frequency questionnaire (FFQ) [25]. Total energy and alcohol intake were estimated using Spanish food composition tables [26].

Main analyses were performed in the evaluable population, including all participants with complete baseline data on exposure, outcomes, and covariates, as well as those who were lost to follow-up at 1, 3, or 6 years.

Baseline characteristics were summarized for the total sample and according to categories of baseline TL. Continuous variables were expressed as means ± standard deviations (SDs), and categorical variables as counts (percentages). Differences across baseline TL categories were examined using one-way analysis of variance (ANOVA) for continuous variables and the chi-square test for categorical variables.

Two-level linear mixed-effects models were fitted to assess the association between baseline TL, modeled as a continuous variable and by predefined categories (< median and ≥ median), and trajectories of ALM or ALMI over 1, 3, and 6 years of follow-up with random intercepts at the household cluster level (as couples from the same household were randomized together), and at the individual participant levels (Fig. 1D). In both cases, an interaction term between baseline TL and time, age (years), sex (male or female), recruiting center (Reus or Pamplona), intervention group (control or intervention), education level (primary or less, secondary, or college), marital status (single, divorced or separated, married, or widower), smoking status (former or never), type 2 diabetes prevalence (yes or no), hypertension prevalence (yes or no), hypercholesterolemia prevalence (yes or no), physical activity (metabolic equivalents in minutes per day), sedentary time (hours per day), alcohol consumption in grams per day (and adding the quadratic term), and total energy intake (kilocalories per day) were included as fixed effects in the multivariable-adjusted models. Results from the linear mixed models were reported as intergroup mean differences with their corresponding 95% confidence intervals (CIs).

A priori interaction analyses were conducted to test for potential effect modification in the association between baseline TL (as continuous) and ALM across baseline categories of age (< 65 vs. ≥ 65 years), sex (male or female), intervention group (control/intervention), type 2 diabetes prevalence (yes or no), smoking status (former or never), physical activity (< median vs. ≥ median), sedentary time (< median vs. ≥ median), alcohol consumption (< median vs. ≥ median), and energy intake (< median vs. ≥ median). Likelihood ratio test was used to evaluate the statistical significance of the interaction terms.

Sensitivity analyses were conducted by performing a complete-case analysis that excluded participants with missing ALM data at 1, 3, and 6 years of follow-up.

All statistical analyses were conducted using Stata/SE version 14.2 (StataCorp LLC, College Station, TX, USA) based on the PREDIMED-Plus study dataset updated to December 19, 2023. All graphs were plotted using GraphPad Prism software v.9.0 (GraphPad Software, San Diego, CA, USA). Statistical significance was defined as a two-tailed P-value < 0.05.