Leukocyte telomere length is inversely associated with a metabolic risk score in Mesoamerican children
Participants in the Nine Mesoamerican Countries Metabolic Syndrome Study (NiMeCoMeS) Group are mentioned in the Acknowledgments section.
Funding information: National Heart, Lung, and Blood Institute, Grant/Award Number: HHSN268200900028C
Abstract
Objective
Leukocyte telomere length (LTL) may be involved in the etiology of the metabolic syndrome (MetS). We examined the associations of LTL with MetS and its components among Mesoamerican children and their adult parents, in a region where MetS prevalence is high.
Methods
We conducted a cross-sectional study of 151 children aged 7–12 years and 346 parents from the capitals of Belize, Honduras, Nicaragua, Costa Rica, Panama, and Chiapas State, Mexico. We quantified LTL by qPCR on DNA extracted from whole blood. In children, we created an age- and sex-standardized metabolic risk score using waist circumference (WC), the homeostasis model of insulin resistance (HOMA-IR), blood pressure, serum high-density lipoprotein (HDL) cholesterol, and serum triglycerides. In adults, MetS was defined according to the National Cholesterol Education Program's Adult Treatment Panel III definition. We estimated mean differences in metabolic risk score and prevalence ratios of MetS across quartiles of LTL using multivariable-adjusted linear and Poisson regression models, respectively.
Results
In children, every 1 LTL z-score was related to an adjusted 0.05 units lower (95% CI: −0.09, −0.02, P = 0.005) MetS risk score, through WC, HOMA-IR, and HDL. Among adults, LTL was not associated with MetS prevalence; however, every 1 LTL z-score was associated with an adjusted 34% lower prevalence of high fasting glucose (95% CI: 3%, 55%, p = .03).
Conclusions
Among Mesoamerican children, LTL is associated with an improved metabolic profile; among adults, LTL is inversely associated with the prevalence of high fasting glucose.
1 INTRODUCTION
The metabolic syndrome (MetS) consists of a set of related cardiometabolic risk factors that strongly predict the incidence of cardiovascular disease, type-2 diabetes (T2D), and mortality (Grundy, 2008). In Latin America, the prevalence of MetS among adults is higher than in many high-income countries (Márquez-Sandoval et al., 2011). MetS components can develop in childhood (Zimmet et al., 2007), and evidence from Latin American and other low- and middle-income countries suggests that MetS has become highly prevalent among children (Caceres et al., 2008; Kelishadi, 2007), partly due to the childhood obesity epidemic in this region (Rivera et al., 2014). For example, in Mexican children aged 6–12 years, 44% of obese children had MetS, and more than one third of normal-weight children had ≥1 MetS component (Guzmán-Guzmán et al., 2015). Children with MetS exhibit signs of early atherosclerosis (Rumińska et al., 2016), and childhood MetS is a strong predictor of both cardiovascular disease (Morrison et al., 2007) and T2D (Morrison et al., 2008) later in life. Hence, identifying modifiable predictors of metabolic dysfunction is critical to address the growing epidemic of chronic disease.
Telomeres are repeated DNA sequences at the ends of chromosomes that protect against loss of genomic DNA during replication and cell division. Leukocyte telomere length (LTL) in humans decreases with chronological age, inflammation, and damage to DNA from oxidative stress (Aviv, 2009). LTL is inversely associated with the risk of cardiovascular disease (Haycock et al., 2014) and diabetes (Zhao et al., 2013), and likely plays a role in the pathogenesis of cancer (Shay, 2016). In addition, shorter LTL has been linked with higher mortality risk in many cohort studies (Arbeev et al., 2020; Wang et al., 2018). Thus, LTL may be a useful biomarker of biological age and risk of chronic disease, particularly cardiometabolic conditions. In adults, LTL has been positively associated with metabolic health (Chen et al., 2009; Demissie et al., 2006; Harte et al., 2012; Huzen et al., 2014; Nordfjäll et al., 2008; Révész et al., 2014; Révész et al., 2015; Révész et al., 2016), although associations with obesity are less consistent (Tzanetakou et al., 2012). In children, LTL may be related to individual MetS components (Al-Attas et al., 2010; Chen et al., 2009), but studies of LTL and MetS are lacking.
The link between LTL and metabolic health is complex; available evidence indicates that their causal relation is likely bidirectional (Kirchner et al., 2017). In light of the increasing burden of MetS, and because most prior studies of LTL and MetS have focused on MetS as the exposure, we aimed to test the causal scenario in which LTL may affect MetS. We conducted a cross-sectional study among families from nine countries in Mesoamerica to examine associations between LTL and MetS among school-aged children and their adult parents in an under-studied population where the prevalence of MetS is high (Villamor et al., 2017).
2 METHODS
2.1 Study design and population
This investigation was conducted in the context of the Nine Mesoamerican Countries Metabolic Syndrome (NiMeCoMeS) Study, a cross-sectional study of school-aged children and their parents from the capital cities of Guatemala, El Salvador, the Dominican Republic, Honduras, Nicaragua, Panama, Costa Rica, and Belize, and the city of Tuxtla Gutiérrez in Chiapas, Mexico. Participants were recruited between 2011 and 2013 in family groups (n = 267) that consisted of one child between ages 7 and 12 years and both biological parents. Additional criteria for enrollment involved not being pregnant or having a pregnant mother, and not having a sibling already invited to participate.
The Institutional Review Boards (IRB) of collaborating institutions in each of the nine countries and the University of Michigan Health and Behavioral Sciences IRB approved all study procedures. All adults provided written informed consent to participate for themselves and their children. Children confirmed their agreement to participate before enrollment.
2.2 Data collection
Recruitment and study procedures have been described in detail elsewhere (Villamor et al., 2017). Briefly, children were recruited at primary schools and invited to attend a research visit with their parents at a study clinic. At this visit, participants completed questionnaires that inquired on sociodemographic characteristics, including their age, education level, smoking status, and indicators of socioeconomic status (SES) such as the number of household assets and household food security status per the Latin American and Caribbean Food Security Scale (ELCSA) (Pérez-Escamilla et al., 2007). Trained study personnel obtained measurements of height, weight, and waist circumference using standardized procedures and calibrated instruments. Waist circumference was measured to the nearest millimeter at the end of an unforced exhalation using an inelastic measuring tape. In adults, it was measured at the midpoint of the lower end of the ribcage and the iliac crest; in children, it was measured above the uppermost lateral border of the right ilium. All anthropometric measures were obtained in triplicate. Blood pressure was measured in triplicate while seated, with at least 1 min between each reading, and the final blood pressure value was calculated as the average of the second and third readings. Finally, a fasting blood sample was collected by antecubital venipuncture. On the day of collection, samples were transported on ice to laboratories in the participants' country for processing and cryostorage at −20°C. Blood samples were then transported frozen to the Institute of Nutrition of Central America and Panama in Guatemala City, Guatemala, where they were stored at −70°C prior to biochemical analysis. An aliquot of whole blood was transported frozen in one batch to the University of Michigan where it was stored at −70°C until analysis. The samples underwent only one thawing cycle for DNA extraction, under homogeneous conditions. Due to logistical constraints, samples for DNA extraction were only collected in participants from Belize, Honduras, Nicaragua, Costa Rica, Panama, and Mexico.
2.3 Laboratory methods
Serum insulin was measured using chemiluminescent immunoassay with an Immulite 1000 system (Siemens Healthcare Diagnostics Products, Tarrytown, NY); plasma glucose concentrations were quantified using an automated chemistry analyzer (Cobas c111 system; Roche Diagnostics, Mannheim, Germany); and serum lipid profiles were determined using enzymatic colorimetric assays on the same system.
We extracted DNA from whole blood using Qiagen Qiasymphony and QIAsymphony DSP DNA Midi Kits (Qiagen, Valencia, CA). LTL was determined at the University of Michigan using a monochrome multiplex qPCR procedure that has been described in detail elsewhere (Cawthon, 2009; Flannagan et al., 2017); an in-depth description of these methods per the Telomere Research Network recommendations (Minimum Reporting Recommendations for PCR-based Telomere Length Measurement, 2019) is also included as Appendix S1. Briefly, all qPCR reactions were carried out using 10 ng of input DNA, iQ SYBR Green Supermix (Bio-Rad Laboratories, Inc., Hercules, CA), and two sets of primers (Invitrogen, Carlsbad, CA, sequences described by Cawthon [Cawthon, 2009]): telg/telc (to amplify telomeres) and albu/albd (to amplify albumin, a single-copy gene). The same batch of standard DNA was serially diluted to make two standard curves on each plate, one for telomeres and one for albumin, with seven points per curve. These curves were used to quantify the telomere (T) and albumin (S) signals for each reaction, and LTL was calculated as the T/S ratio. All samples were run in triplicate, and all plates included a positive control sample to assess inter-assay variation, as well as a no template control to detect spurious amplification. We calculated the intraclass correlation coefficient of the three replicates (ICC = 0.92) using a one-way, random-effects, absolute agreement model (McGraw & Wong, 1996). Intra- and inter-assay coefficients of variation were 16% and 37%, respectively. To further document the quality of our LTL measurements, we tested a random set of 47 samples at Dr. Elizabeth Blackburn's reference laboratory at the University of California San Francisco. The correlation between T/S values obtained at both labs was very high (r = 0.88, Figure S1). LTL data also had high internal validity; for example, child values' correlation with their mother's (r = 0.25) was within the expected range (Broer et al., 2013; Eisenberg, 2014). Moreover, associations with sex (Table S1) were in the expected direction per existing literature (Ly et al., 2019; Müezzinler et al., 2013).
2.4 Definition of exposure
We estimated the median of the three LTL replicate measurements per participant (Villamor & Bosch, 2015) to use as the exposure variable in the analyses. We then transformed LTL values to z-scores, as recommended (Verhulst, 2020), separately for children and adults using each subgroup as the corresponding standard. LTL z-scores were categorized into quartiles and also considered as a continuous variable.
2.5 Definition of outcomes
2.5.1 Children
Because the metabolic syndrome is undefined in children <10 years old, we estimated a continuous metabolic risk score using values on the same criteria used in adults, as previously recommended (Eisenmann, 2008). We used the homeostasis model assessment of insulin resistance (HOMA-IR) as the measure of insulin resistance and mean arterial pressure (MAP) as the blood pressure criterion. We first obtained sex- and age-standardized scores for WC, HOMA-IR, MAP, high-density lipoprotein (HDL) cholesterol and serum triglycerides (TAG) by performing the regression of the natural logarithmic transformation of each of these variables on sex and log-transformed age with the use of linear regression models. The standardized regression residuals for WC, HOMA-IR, MAP and TAG, plus the HDL cholesterol residual multiplied by −1, were then averaged to create a metabolic risk score, where higher values reflect a worse metabolic profile.
2.5.2 Adults
The primary outcome of interest was the presence of MetS according to the National Cholesterol Education Program's ATP III criteria. Abdominal obesity was defined as WC > 102 cm in men or WC >88 cm in women; high fasting glucose as ≥100 mg/dl; high blood pressure as systolic blood pressure ≥ 130 mmHg, diastolic blood pressure ≥ 85 mmHg or treatment with an antihypertensive drug; low HDL cholesterol as serum concentration < 40 mg/dl in men or < 50 mg/dl in women, or drug treatment for reduced HDL cholesterol; and high TAG as serum concentration ≥ 150 mg/dl or drug treatment for elevated TAG. MetS was defined as the presence of three or more of these criteria.
2.6 Covariates
In children, we estimated sex-specific height-for-age z-scores according to the World Health Organization reference (de Onis et al., 2007). Parental smoking was defined as ever smoking by one or both parents. In adults, smoking status was classified as never, past, or current. The number of household assets was determined as the sum of affirmative responses to owning a car, bicycle, refrigerator/freezer, gas stove, electric stove, blender, microwave, washing machine, color television, sound set, computer, or Internet. The presence of severe food insecurity in the household was defined as a sum of affirmative answers ≥11 to the 16 questions on food security experiences over the prior 3 months in the ELCSA survey.
2.7 Data analyses
There were 151 children and 346 adults with non-missing data on both LTL and metabolic syndrome outcomes and these constituted the analytic sample. All analyses were conducted separately for children and adults.
2.7.1 Children
We compared the distribution of the metabolic risk score overall and for each component across quartiles of LTL z-score using means and standard deviations. Next, we estimated mean metabolic risk score differences with 95% CI between LTL z-score quartiles using bivariate linear regression. We then fit two adjusted models, one with country of origin which was the largest source of variability in LTL in this population (Flannagan et al., 2017) and a multivariable-adjusted model involving additional independent predictors of the metabolic syndrome (Villamor et al., 2017) or correlates of LTL that are not affected by it (to avoid over-adjustment), such as parental MetS and parental LTL. In addition to country, these included the presence of metabolic syndrome in the parents, child's height-for-age z-score, mother's height and LTL z-score, father's height and LTL z-score, and parental smoking. We also adjusted for SES by including the number of household assets as a continuous variable, as well as an indicator variable for severe food insecurity. The metabolic risk score was standardized for sex and age using the study population as the reference; therefore, estimates of association are intrinsically adjusted for these variables. Because the multivariable-adjusted associations between LTL and metabolic risk score seemed to follow a dose–response relation, we also estimated the difference in metabolic risk score per 1 z-score difference in LTL.
2.7.2 Adults
In bivariate analyses, we compared the prevalence of MetS and its components across quartiles of LTL z-score. Next, we estimated prevalence ratios with 95% CI. These were first adjusted for country alone and subsequently for independent predictors of MetS or its components plus correlates of LTL which included sex, age and height, and smoking status (Flannagan et al., 2017), as well as SES, as indicated by household assets and food insecurity. Robust estimates of variance were specified in all models to account for within-family correlations among adults. All analyses were performed using SAS version 9.4.
3 RESULTS
3.1 Children
Mean ± SD age of children was 9.9 ± 1.6 years; 53% were girls (Table S2). Mean ± SD LTL and metabolic risk score were 3.32 ± 1.05 and − 0.01 ± 0.23, respectively. In bivariate analysis, LTL was not associated with the metabolic risk score (Table 1); however, after adjustment for country, LTL was inversely related to this score. In multivariable analysis, every 1 LTL z-score was associated with an adjusted mean 0.05 units lower risk score (p = .005) (Table 1). LTL was inversely associated with the WC score (−0.03 adjusted mean units per 1 LTL z-score, p = .01) and the HOMA-IR score (−0.12 adjusted mean units per 1 LTL z-score, p = .03), and positively associated with the serum HDL score (0.06 adjusted mean units per 1 LTL z-score, p = .002). LTL was not significantly related to the MAP or the serum TAG scores. Regression coefficients for adjustment covariates are included in Table S3.
| Metabolic syndrome score | LTL z-score quartile (Q) | Mean difference (95% CI) per 1 LTL z-scorea | p | |||
|---|---|---|---|---|---|---|
| Q1 n = 37 | Q2 n = 38 | Q3 n = 38 | Q4 n = 38 | |||
| Median LTL | 2.08 | 2.93 | 3.62 | 4.53 | ||
| Overall metabolic risk score | ||||||
| Mean ± SD | −0.06 ± 0.26 | 0.06 ± 0.24 | −0.01 ± 0.19 | −0.03 ± 0.20 | ||
| Unadjusted differences (95% CI) | Reference | 0.12 (0.01, 0.23) | 0.05 (−0.05, 0.16) | 0.03 (−0.07, 0.14) | 0.00 (−0.04, 0.03) | .90 |
| Country-adjusted differences | Reference | 0.05 (−0.09, 0.19) | −0.01 (−0.15, 0.12) | −0.05 (−0.19, 0.09) | −0.04 (−0.08, 0.00) | .03 |
| Multivariable-adjusted differencesb | Reference | 0.02 (−0.12, 0.16) | −0.04 (−0.17, 0.09) | −0.09 (−0.22, 0.05) | −0.05 (−0.09, −0.02) | .005 |
| Waist circumference score | ||||||
| Mean ± SD | −0.03 ± 0.14 | 0.04 ± 0.16 | 0.01 ± 0.14 | −0.02 ± 0.15 | ||
| Unadjusted differences (95% CI) | Reference | 0.06 (0.00, 0.13) | 0.03 (−0.03, 0.09) | 0.01 (−0.06, 0.07) | 0.00 (−0.03, 0.02) | .83 |
| Country-adjusted differences | Reference | 0.03 (−0.04, 0.10) | 0.00 (−0.07, 0.06) | −0.04 (−0.11, 0.03) | −0.02 (−0.05, 0.00) | .08 |
| Multivariable-adjusted differencesb | Reference | 0.01 (−0.05, 0.07) | −0.02 (−0.08, 0.04) | −0.04 (−0.10, 0.01) | −0.03 (−0.05, −0.01) | .01 |
| HOMA-IR score | ||||||
| Mean ± SD | −0.13 ± 0.71 | 0.17 ± 0.63 | −0.02 ± 0.52 | −0.02 ± 0.55 | ||
| Unadjusted differences (95% CI) | Reference | 0.30 (0.00, 0.60) | 0.11 (−0.17, 0.39) | 0.11 (−0.18, 0.39) | 0.01 (−0.09, 0.10) | .89 |
| Country-adjusted differences | Reference | 0.13 (−0.34, 0.60) | −0.06 (−0.49, 0.37) | −0.10 (−0.54, 0.35) | −0.09 (−0.20, 0.02) | .12 |
| Multivariable-adjusted differencesb | Reference | 0.03 (−0.41, 0.48) | −0.13 (−0.54, 0.27) | −0.20 (−0.63, 0.22) | −0.12 (−0.22, −0.01) | .03 |
| MAP score | ||||||
| Mean ± SD | 0.00 ± 0.14 | 0.04 ± 0.18 | 0.02 ± 0.12 | 0.03 ± 0.16 | ||
| Unadjusted differences (95% CI) | Reference | 0.04 (−0.03, 0.11) | 0.02 (−0.04, 0.08) | 0.03 (−0.04, 0.10) | 0.00 (−0.02, 0.03) | .93 |
| Country-adjusted differences | Reference | 0.03 (−0.07, 0.12) | −0.01 (−0.10, 0.08) | −0.01 (−0.11, 0.09) | −0.02 (−0.05, 0.01) | .22 |
| Multivariable-adjusted differencesb | Reference | 0.03 (−0.06, 0.11) | 0.00 (−0.08, 0.07) | 0.00 (−0.08, 0.08) | −0.01 (−0.04, 0.01) | .34 |
| Serum HDL score | ||||||
| Mean ± SD | 0.09 ± 0.22 | −0.01 ± 0.23 | 0.04 ± 0.24 | 0.10 ± 0.27 | ||
| Unadjusted differences (95% CI) | Reference | −0.10 (−0.21, 0.00) | −0.05 (−0.16, 0.05) | 0.01 (−0.10, 0.12) | 0.02 (−0.02, 0.05) | .40 |
| Country-adjusted differences | Reference | 0.01 (−0.08, 0.11) | 0.05 (−0.04, 0.14) | 0.11 (−0.01, 0.22) | 0.05 (0.01, 0.10) | .01 |
| Multivariable-adjusted differencesb | Reference | 0.01 (−0.08, 0.10) | 0.07 (−0.01, 0.16) | 0.13 (0.02, 0.23) | 0.06 (0.02, 0.10) | .002 |
| Serum triglycerides score | ||||||
| Mean ± SD | −0.06 ± 0.40 | 0.04 ± 0.41 | 0.00 ± 0.36 | −0.03 ± 0.37 | ||
| Unadjusted differences (95% CI) | Reference | 0.10 (−0.08, 0.28) | 0.06 (−0.11, 0.24) | 0.04 (−0.14, 0.21) | 0.00 (−0.06, 0.05) | .97 |
| Country-adjusted differences | Reference | 0.09 (−0.12, 0.30) | 0.05 (−0.16, 0.26) | 0.00 (−0.22, 0.23) | −0.03 (−0.10, 0.03) | .34 |
| Multivariable-adjusted differencesb | Reference | 0.04 (−0.18, 0.25) | 0.02 (−0.18, 0.22) | −0.06 (−0.28, 0.17) | −0.04 (−0.11, 0.03) | .24 |
- Abbreviations: LTL, leukocyte telomere length; HOMA-IR, homeostasis model assessment of insulin resistance; MAP, mean arterial pressure; HDL, high density lipoprotein; CI, confidence interval.
- a From a linear regression model with LTL z-score as a continuous predictor.
- b From a linear regression model with each score as a continuous outcome, adjusted for country, presence of metabolic syndrome in the parents (two indicators), child's height-for-age z-score, mother's height and LTL z-score, father's height and LTL z-score (continuous), parental smoking (two indicators), and socioeconomic status (household assets and an indicator for severe food insecurity). Robust variances were specified in all models.
3.2 Adults
Mean ± SD age of adults was 38.7 ± 7.6 years; 50.3% were women (Table S4). Mean ± SD LTL was 2.20 ± 0.80; the prevalence of MetS was 35.6%. LTL was not significantly associated with the MetS (Table 2); however, there was a linear, inverse relation between LTL and high fasting glucose in bivariate and multivariable analyses. Every 1 LTL z-score was associated with an adjusted 34% lower prevalence of high fasting glucose (p = .03). LTL was not significantly related to other metabolic syndrome components. Parameter estimates for adjustment covariates are presented in Table S5.
| Metabolic syndrome component | LTL z-score quartile (Q) | Prevalence ratio (95% CI) per 1 LTL z-scorea | p | |||
|---|---|---|---|---|---|---|
| Q1 n = 86 | Q2 n = 87 | Q3 n = 86 | Q4 n = 87 | |||
| Median LTL | 1.32 | 1.86 | 2.33 | 3.13 | ||
| Metabolic syndrome | ||||||
| Prevalence, % | 31.4 | 40.2 | 39.5 | 31.0 | ||
| Unadjusted prevalence ratio (95% CI) | 1.00 | 1.28 (0.86, 1.90) | 1.26 (0.84, 1.89) | 0.99 (0.64, 1.53) | 0.97 (0.84, 1.11) | .65 |
| Country-adjusted prevalence ratio | 1.00 | 1.30 (0.88, 1.93) | 1.23 (0.82, 1.87) | 1.01 (0.65, 1.57) | 0.98 (0.84, 1.13) | .76 |
| Multivariable-adjusted prevalence ratiob | 1.00 | 0.83 (0.55, 1.26) | 1.06 (0.73, 1.54) | 0.79 (0.52, 1.20) | 0.99 (0.85, 1.15) | .88 |
| Abdominal obesity | ||||||
| Prevalence, % | 50.0 | 49.4 | 47.7 | 44.8 | ||
| Unadjusted prevalence ratio (95% CI) | 1.00 | 0.99 (0.73, 1.33) | 0.95 (0.70, 1.30) | 0.90 (0.64, 1.25) | 0.96 (0.86, 1.08) | .50 |
| Country-adjusted prevalence ratio | 1.00 | 1.03 (0.77, 1.37) | 0.98 (0.71, 1.34) | 0.91 (0.66, 1.27) | 0.97 (0.86, 1.09) | .63 |
| Multivariable-adjusted prevalence ratiob | 1.00 | 0.86 (0.65, 1.14) | 1.04 (0.81, 1.33) | 1.07 (0.80, 1.43) | 0.92 (0.82, 1.04) | .17 |
| High fasting glucose | ||||||
| Prevalence, % | 11.6 | 10.3 | 9.3 | 3.5 | ||
| Unadjusted prevalence ratio (95% CI) | 1.00 | 0.89 (0.38, 2.08) | 0.80 (0.34, 1.90) | 0.30 (0.08, 1.05) | 0.65 (0.45, 0.95) | .02 |
| Country-adjusted prevalence ratio | 1.00 | 0.87 (0.37, 2.05) | 0.78 (0.34, 1.82) | 0.30 (0.08, 1.07) | 0.65 (0.44, 0.96) | .03 |
| Multivariable-adjusted prevalence ratiob | 1.00 | 0.36 (0.10, 1.33) | 1.07 (0.42, 2.74) | 1.18 (0.50, 2.83) | 0.66 (0.45, 0.97) | .03 |
| High blood pressure | ||||||
| Prevalence, % | 20.9 | 28.7 | 19.8 | 20.9 | ||
| Unadjusted prevalence ratio (95% CI) | 1.00 | 1.37 (0.82, 2.29) | 0.94 (0.51, 1.74) | 1.00 (0.57, 1.77) | 0.95 (0.79, 1.14) | .58 |
| Country-adjusted prevalence ratio | 1.00 | 1.37 (0.82, 2.28) | 0.90 (0.50, 1.64) | 0.94 (0.53, 1.66) | 0.92 (0.76, 1.11) | .38 |
| Multivariable-adjusted prevalence ratiob | 1.00 | 1.17 (0.65, 2.13) | 1.56 (0.88, 2.76) | 1.00 (0.54, 1.83) | 0.99 (0.82, 1.21) | .94 |
| Low HDL cholesterol | ||||||
| Prevalence, % | 75.6 | 80.5 | 88.4 | 79.3 | ||
| Unadjusted prevalence ratio (95% CI) | 1.00 | 1.06 (0.92, 1.23) | 1.17 (1.02, 1.34) | 1.05 (0.90, 1.23) | 0.99 (0.93, 1.06) | .81 |
| Country-adjusted prevalence ratio | 1.00 | 1.07 (0.93, 1.24) | 1.18 (1.03, 1.35) | 1.09 (0.93, 1.28) | 1.01 (0.94, 1.08) | .85 |
| Multivariable-adjusted prevalence ratiob | 1.00 | 0.92 (0.81, 1.05) | 0.92 (0.82, 1.04) | 0.86 (0.75, 1.00) | 1.00 (0.93, 1.07) | .94 |
| High triglycerides | ||||||
| Prevalence, % | 44.2 | 52.9 | 58.1 | 48.3 | ||
| Unadjusted prevalence ratio (95% CI) | 1.00 | 1.20 (0.86, 1.67) | 1.32 (0.97, 1.78) | 1.09 (0.78, 1.53) | 0.99 (0.89, 1.11) | .90 |
| Country-adjusted prevalence ratio | 1.00 | 1.18 (0.85, 1.65) | 1.29 (0.95, 1.75) | 1.12 (0.79, 1.57) | 1.00 (0.89, 1.12) | .99 |
| Multivariable-adjusted prevalence ratiob | 1.00 | 0.94 (0.71, 1.24) | 0.90 (0.69, 1.17) | 0.77 (0.57, 1.03) | 1.04 (0.93, 1.17) | .46 |
- Abbreviations: LTL, leukocyte telomere length; HDL, high density lipoprotein; CI, confidence interval.
- a From a Poisson regression model with LTL z-score as a continuous predictor.
- b From a Poisson regression model with metabolic syndrome or each component as a dichotomous outcome, adjusted for country, sex, age and height (continuous), smoking status (indicators for past or current), and socioeconomic status (household assets and an indicator for severe food insecurity). Robust variances were specified in all models.
4 DISCUSSION
In this cross-sectional study of Mesoamerican families, we found that LTL was inversely related to a MetS risk score in children, mainly through inverse associations with WC and HOMA-IR, and a positive association with serum HDL cholesterol. Among adults, LTL was inversely related to the prevalence of high fasting glucose in a dose–response manner.
Few studies have explored the association between LTL and metabolic health among children, although some have reported inverse associations between childhood obesity, a component of MetS, and LTL (Buxton et al., 2011; Clemente et al., 2019; Lamprokostopoulou et al., 2019). In a lifestyle intervention among overweight and obese adolescents, those with higher LTL at baseline had a larger decrease in body weight (García-Calzón et al., 2014). Meanwhile, in a randomized trial of a different lifestyle intervention in children with abdominal obesity, higher baseline LTL predicted larger decreases in blood glucose during the intervention (Morell-Azanza et al., 2020), in line with our findings for HOMA-IR. In the Bogalusa heart study, higher childhood HDL was associated with higher adult LTL (Chen et al., 2009), which is consistent with the positive association between LTL and HDL in our study, although we did not follow children into adulthood. In a cross-sectional study of children in middle childhood, WC was the strongest predictor of LTL in girls, whereas systolic blood pressure was the most important predictor in boys (Al-Attas et al., 2010). These results are consistent with the inverse relation between LTL and WC in Mesoamerican children; however, we did not find an association with MAP. Differences in study populations, as well as adjustment approaches, might explain these discrepancies.
In adults, most studies have reported a positive relation between LTL and metabolic health. Cross-sectional studies have found associations between shorter LTL and elevated WC (Nordfjäll et al., 2008), TAG (Harte et al., 2012), insulin resistance (Demissie et al., 2006), and hypertension (Demissie et al., 2006). In the Netherlands Study of Depression and Anxiety, shorter baseline LTL was associated with an adverse metabolic profile across all MetS components at baseline, 2 years, and 6 years (Révész et al., 2014). In the same cohort, an adverse baseline profile of MetS components (high WC, elevated glucose, low HDL, and total number of metabolic disregulations) was associated with shorter LTL at baseline and follow-up; also, greater 6-year increase in WC was associated with a greater decrease in 6-year LTL (Révész et al., 2015). In the PREVEND study, higher baseline plasma glucose and waist-to-hip ratio, as well as lower HDL, were associated with greater LTL attrition over follow up (Huzen et al., 2014). More recently, results from the CARDIA study found that HDL was positively associated with LTL during follow-up, and 10-year increase in WC was related to greater telomere attrition (Révész et al., 2018). Finally, in a prospective twin study, baseline LTL was inversely associated with changes in insulin resistance over an average of 12 years of follow-up (Verhulst et al., 2016), in line with our finding of an inverse association between LTL and prevalence of high fasting glucose in adults. Unlike many prior studies, we did not find associations between LTL and other MetS components; however, differences in study design and participant characteristics could be responsible for these divergent results.
Mechanisms linking LTL and metabolic health remain incompletely understood. It has been hypothesized that inflammation and oxidative stress are responsible for associations between LTL and insulin resistance (Gardner et al., 2005). These factors may also explain the connection between LTL and HDL cholesterol, as HDL reduces oxidation and inflammation (Chen et al., 2009). However, other authors have noted that, because relative LTL is largely determined prior to adulthood, metabolic outcomes in adults are likely preceded by differences in LTL, as opposed to causing those differences (Verhulst et al., 2016). Evidence from animal models supports a potential impact of LTL on metabolic health. In mice, short telomeres impair beta cell signaling, contributing to hyperglycemia (Guo et al., 2011), and promote mitochondrial dysfunction, which leads to a number of downstream adverse metabolic effects (Sahin et al., 2011). In these models, disruption of a protein that is part of a protective complex at telomeres results in metabolic dysregulation with effects that include glucose intolerance, insulin resistance, and accumulation of abdominal fat (Martínez et al., 2013; Yeung et al., 2013). Ongoing studies will be needed to investigate the extent to which LTL may be a cause or consequence of metabolic health.
Our study has several strengths. The relation between metabolic syndrome and LTL has not been studied in children; thus, our results address an important gap, which is crucial in developing early interventions to improve metabolic health. In addition, our population consisted of people from a region where childhood metabolic syndrome and obesity represent a substantial public health burden. Because we studied families of adult parents and their children, we were also able to compare the LTL-metabolic syndrome associations between children and adults, and we had the opportunity to adjust the results in children for parental LTL.
Some limitations should also be noted. Because this study was cross-sectional, we cannot determine the direction of the association between LTL and MetS. As discussed previously, it is possible that telomere shortening disrupts metabolic function, that metabolic health affects LTL, or that the association is bidirectional. We also cannot discard the possibility of residual confounding, which results in non-causal estimates of association due to the effects of common causes of exposure (LTL) and outcome (MetS) (Bateson & Nettle, 2018; Lash et al., 2021). One example of an unmeasured confounder for which we could not control is MetS during pregnancy; this has been related to shorter telomeres later in childhood (McAninch et al., 2020) and could be a cause of MetS in children through shared genetic and environmental factors. Although we were able to adjust for parental MetS during childhood, it is possible that MetS in pregnancy could introduce additional confounding that we were not able to capture. Measurement error in the ascertainment of exposure, outcome, and covariates is another potential limitation of the study. Furthermore, associations between LTL and metabolic syndrome cannot be quantitatively compared between children and adults because the outcomes are in different scales. Last, generalizability may be limited to populations with comparable distributions of LTL and metabolic outcomes.
In conclusion, higher LTL is associated with a more favorable metabolic profile among Mesoamerican children, specifically through lower WC and HOMA-IR, and higher HDL. In adults, higher LTL is associated with a lower prevalence of high fasting glucose. Prospective investigations of these questions are warranted, especially in children, to enhance strategies for improving metabolic health throughout life.
ACKNOWLEDGMENTS
The study was funded by the United States National Heart, Lung, and Blood Institute, grant HHSN268200900028C.
Participants in the Nine Mesoamerican Countries Metabolic Syndrome Study (NiMeCoMeS) Group:
Mexico: Erika Lopez, Liz Peña, Alejandra Maldonado, Aldeni Vasquez, Aldrin Lopez.
Belize: Lilly Mahung, Diomar Salazar.
Guatemala: Ana Victoria Román, Fernanda Kroker, Maria Alejandra Cordova, Regina Garcia, Lilian Navas.
El Salvador: Josefina Sibrian, Mauricio Flores, Noel Avalos.
Honduras: Astarte Alegria, Jorge A. Sierra, Hector Murillo.
Nicaragua: Ana María Gutierrez, Carmen María Flores, Mario Romero.
Costa Rica: Emilce Ulate, Natalia Valverde, Andrea Fiatt, Juan Manuel Valverde.
Panama: Flavia Fontes, Raisa Rodriguez, Emerita Pons, Lino Chue, Elka Gonzalez.
Dominican Republic: Rafael Montero, Francisco Torres, Amarilis Then, Melvi Perez.
CONFLICT OF INTEREST
The authors declare that they have no competing interests.
AUTHOR CONTRIBUTIONS
Joshua Garfein drafted the manuscript. Kerry S. Flannagan optimized the telomere measurement protocol and quantified LTL. Kerry S. Flannagan and Eduardo Villamor conceptualized the study. Danielle Rittman, Eduardo Villamor, and Joshua Garfein performed the data analysis. Manuel Ramirez-Zea oversaw data collection. Manuel Ramirez-Zea and Eduardo Villamor designed the Nine Mesoamerican Countries Metabolic Syndrome Study and obtained funding. All authors read and approved the final version of the manuscript.
Open Research
DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.
